A HYBRID RECOMMENDATION SYSTEM FOR RECOMMENDING PRODUCT
ADVERTISEMENTS
TECHNICAL FIELD
[001]
The present invention relates to the field of electronic advertising and, in particular,
relates to the recommendation of products through one or more advertisements to one or
more users.
BACKGROUND
[002]
With the advent of the internet, an escalating amount of information is available on
the web. Nowadays, internet searching has become one of the top most tasks performed by
users for searching products or items related to their interest. In the current scenario, this
escalating amount of information is being utilized for advertising purposes and targeting
users who are looking for one or more products on various publisher websites. Moreover, a
large number of users are interested in searching, viewing and buying one or more products
online through one or more e-commerce publishers. The one or more e-commerce publisher
includes Flipkart, Jabong, Snapdeal, Amazon and the like. For example, a user may be
looking for a new smart phone or another user may be interested in buying footwear or
apparels.
[003]
In addition, one or more publishers nowadays employ one or more recommendation
systems for effective targeting of the users for the advertising purposes. The one or more
publishers are able to target the users for showing advertisements related to the products of
the interests of the users. Further, each user is looking for a different type of attributes in a
product while browsing on the publisher website. For example, a user may give higher
weightage to a processor speed and RAM while searching for mobile phones or another user
may give higher weightage to size and color while searching for mobile phones. The current
recommendation systems utilize this information for recommending one or more products
which are similar to the attributes that the users are looking for. The recommendation system
helps in improving an online experience for the users and helps the one or more publishers in
gaining valuable insight regarding the preferences of the users.[004]
Going further, the recommendation systems record the response of the users based
on interaction with the one or more recommended advertisements.
Moreover, the
recommendation systems record the one or more actions taken by one or more users for a
particular product over a period of time. In addition, the recommendation systems calculate a
probability of the users for buying the one or more products. The one or more users are
divided based on their interests in the one or more products. Further, the interests of the one
or more users keep changing over the period of time.
In addition, some of the
recommendation systems perform clustering of the one or more users for effective targeting.
[005]
Moreover, the recommendation systems presently known in the art are build using
either user based collaborative recommendation, item based recommendation, content based
recommendation and demography based recommendation.
Some of the hybrid
recommendation systems have been built using a combination of content and user based
collaborative recommendation algorithm. Several present systems are known in the art
which divides the users in one or more clusters. One such system employs a clustering
process for creating clusters of user using a distance metric. Another such system allows the
classifying of the users based on one or more user characteristics. Yet another system allows
the recommendation of the products from a population of one or more products based on
intelligence contained in processing elements and subjective and/or objective product
information received from consumers or inputs to the systems as part of the initial setup. Yet
another system recommends one or more items by fetching ratings provided to the items by
various users and calculating a preference score based on the item ratings.
[006]
However, the present hybrid recommendation systems known in the art do allow the
clustering of the users based on calculation of an average cluster centroid. In addition, the
present hybrid recommendation systems do not allow recommendation of one or more
products similar to the product currently viewed by the user based on calculation of the
average cluster centroid. Moreover, the present hybrid recommendation systems do not
allow computation of propensity to buy the product for the user based on past interaction of
the one or more users with the product. Further, the present hybrid recommendation systems
do not recommend the product through the advertisements based on the propensity to buy.
Furthermore, the present hybrid recommendation systems do not allow deciding of the
advertiser of one or more advertisers whose recommended product will be displayed to theuser. Moreover, there is no such system known in the art which utilizes a combination of a
user based model, an item based model, a content based model and a demography based
model.
[007]
In the light of the above stated discussion, there is a need for system that overcomes
the above stated disadvantages.
SUMMARY
[008]
In an aspect of the present disclosure, a hybrid recommendation system for
recommending a plurality of advertisements to a user of one or more users viewing one or
more products on a publisher of one or more publishers is provided.
The hybrid
recommendation system includes a collection module in a processor, the collection module is
configured to collect a first set of pre-defined attributes associated with the user of the one or
more users viewing the one or more products on the publisher of the one or more publishers;
a clustering module in the processor, the clustering module is configured to classify the user
of the one or more users in one or more clusters based on the first set of pre-defined
attributes; a determination module in the processor, the determination module is configured
to determine one or more characteristics associated with a product of the one or more
products currently viewed by the user of the one or more users; a gathering module in the
processor, the gathering module is configured to gather data associated with other one or
more products similar to the product currently viewed by the user based on the determination
of the one or more characteristics and the identification of the cluster of the one or more
clusters; a computational module in the processor, the computational module is configured to
compute a weighted average propensity to buy for the user of the one or more users and a
recommendation engine in the processor, the recommendation engine is configured to
recommend the plurality of advertisements associated with the product of the one or more
products currently viewed by the user and the other one or more products having a highest
weighted average propensity to buy. The one or more clusters are created based on a first set
of parameters. The classifying and clustering is done for identifying a cluster of the one or
more clusters to which the user of the one or more users belongs. The identifying is based on
calculation of an average cluster centroid value for the user of the one or more users. The
one or more clusters are created for identifying one or more parameters for the clustering.The clustering is done at pre-defined intervals of time. The determining of the one or more
characteristics is based on the identification of the cluster of the one or more clusters. The
other one or more products depict similar characteristics to the one or more characteristics
associated with the product of the one or more products currently viewed by the user. The
other one or more products belong to the identified cluster of the one or more clusters
corresponding to the product of the one or more products currently viewed by the user. The
weighted average propensity to buy is computed for the product of the one or more products
currently viewed by the user of the one or more users and the other one or more products
similar to the product currently viewed by the user. The weighted average propensity to buy
is computed based on a second set of pre-defined attributes and a pre-defined criterion. The
pre-defined criterion is based on analyzing a past behavior of a set of users of the one or
more users who have bought the product of the one or more products and the other one or
more products and taken one or more actions corresponding to the product of the one or more
products and the other one or more products.
BRIEF DESCRIPTION OF THE FIGURES
[009]
Having thus described the invention in general terms, reference will now be made to
the accompanying drawings, which are not necessarily drawn to scale, and wherein:
[0010]
FIG. 1A and FIG. 1B illustrates a general overview of a system for recommending a
plurality of advertisements, in accordance with various embodiments of the present
disclosure;
[0011]
FIG. 2 illustrates a block diagram of a communication device, in accordance with
various embodiments of the present disclosure;
[0012]
FIG. 3 illustrates an example snapshot showing classification of one or more users,
in accordance with various embodiments of the present disclosure;
[0013]
FIG. 4 illustrates an example snapshot for showing calculation of weighted average
propensity to buy, in accordance with various embodiments of the present disclosure;
[0014]
FIG. 5 illustrates an example snapshot for calculating a recommendation score, in
accordance with various embodiments of the present disclosure;[0015]
FIG. 6A, FIG. 6B and FIG. 6C illustrates an example snapshot for taking a
decision associated with an advertiser of one or more advertisers, in accordance with various
embodiments of the present disclosure;
DETAILED DESCRIPTION
[0016]
It should be noted that the terms "first", "second", and the like, herein do not denote
any order, quantity, or importance, but rather are used to distinguish one element from another.
Further, the terms "a" and "an" herein do not denote a limitation of quantity, but rather denote
the presence of at least one of the referenced item.
[0017]
FIG. 1 illustrates a general overview of the system 100 for recommending a
plurality of advertisements to a user of one or more users viewing one or more products on a
publisher of one or more publishers, in accordance with various embodiments of the present
disclosure. The system 100 includes a communication device 104 associated with a user
102, a communication device 108 associated with a user 106, one or more publishers 110, a
communication network 112 and a hybrid recommendation system 114.
The hybrid
recommendation system 114 is configured to perform the recommending of the plurality of
advertisements to the user of the one or more users who are viewing the one or more
products on the publisher of the one or more publishers.
[0018]
Going further, the user 102-106 may be any person or individual accessing the
corresponding communication device 104-108. In an embodiment of the present disclosure,
the users 102-106 are owners of the corresponding communication device 104 and the
communication device 108.
Moreover, the communication device 104 and the
communication device 108 are portable communication devices.
Examples of the
communication devices 104-108 include but may not be limited to a smart phone, a desktop
computer, a tablet, a laptop, a personal digital assistant or any other electronic portable
device presently known in the art. In addition, each of the communication devices 104-108
is connected to an internet broadband system, a local area network, a wide area network, a
digital or analog cable television network or any other communication network presently
known in the art. The internet broadband system maybe a wired or a wireless system.[0019]
Further, the user 102 accesses a browser of one or more browsers on the
corresponding communication device 104 and the user 106 accesses the browser of the one
or more browsers on the corresponding communication device 108. Furthermore, the user
102 and the user 106 access a website owned by a publisher of the one or more publishers
110 through the browser of the one or more browsers. Examples of the one or more
publishers 110 include but may not be limited to facebook, groupon, flipkart, amazon,
snapdeal, myntra, jabong and mobile applications. Each of the one or more publishers 110
correspond to one or more website owners for providing or displaying content to the users
102-106. In an embodiment of the present disclosure, the one or more publishers 110
correspond to one or more e-commerce publishers selling the one or more products. In an
embodiment of the present disclosure, the users 102-106 want to buy the one or more
products online by accessing the one or more publishers 114. The one or more publishers
110 provide a wide range of the one or more products for the users 102-106 to choose from.
[0020]
Further, the one or more publishers 110 provide space, areas or a part of their web
pages for advertising purposes. These areas or spaces on the web pages are referred to as
advertisement slots. The web page can have the various advertisement slots depending on
choice of each of the one or more publishers 110. The one or more publishers 110 advertise
products, services or businesses to the users 102-104 for generating revenue. It may be
noted that the term publisher in context of the present application may be referred to as
publisher website which may have advertisement slots for advertising.
[0021]
In addition, the content accessed by the user 102 and the user 106 may be any
content related to interests of the user 104 and the user 106. Moreover, the content accessed
by the users 102-106 corresponds to the one or more products associated with the interests
of each of the users 102-106. The one or more products include but may not be limited to
electronic products (mobiles, tablets, computers, laptops and the like), electrical products,
clothing and accessories, (jeans, shirts, tops, watches, sunglasses and the like), household
items (kitchen and home appliances, indoor lighting products, pet supplies, home furnishing
and the like), footwear products (sneakers, slippers, flip flops, formal shoes and the like) and
sports and fitness products (cricket products, football products, exercise products, cycling
products and the like).[0022]
Moreover, the user 102 and the user 106 may access the content through one or more
mobile applications. In an embodiment of the present disclosure, the website or the mobile
application accessed by the users 102-106 on the corresponding communication devices
104-108 may show content related to interests of the user 102-104. For example, the user
102 may be interested in footwear products, the electronic products, the sports and fitness
products, household products and the like.
[0023]
Going further, each of the communication devices 104-108 is associated with the
hybrid recommendation system 114. In an embodiment of the present disclosure, each of
the communication devices 104-108 is associated with the hybrid recommendation system
114 through the communication network 112. In addition, the one or more publishers 110
are associated with the hybrid recommendation system 114. In an embodiment of the
present disclosure, the one or more publishers 110 are associated with the hybrid
recommendation system 114 through the communication network 112. The communication
network 112 enables the hybrid recommendation system 114 to track one or more activities
performed by each of the users 102-106 on the corresponding publisher of the one or more
publishers 110. Moreover, the communication network 112 provides a medium for transfer
of information associated with the users 102-106 to the hybrid recommendation system 114.
[0024]
Further, the medium for communication may be infrared, microwave, radio
frequency (RF) and the like. The communication network 112 include but may not be
limited to a local area network, a metropolitan area network, a wide area network, a virtual
private network, a global area network, a home area network or any other communication
network presently known in the art. The communication network 112 is a structure of
various nodes or communication devices connected to each other through a network
topology method.
Examples of the network topology include a bus topology, a star
topology, a mesh topology and the like.
[0025]
Going further, the hybrid recommendation system 114 is a system for recommending
the plurality of advertisements to the user 102 and the user 106 on the corresponding
publisher of the one or more publishers 110 when the users 102-106 view the one or more
products and take one or more actions. Moreover, the hybrid recommendation system 114
records the one or more activities performed by the users 102-106 and recommend the one
or more products through the plurality of advertisements based on the one or more activities.The one or more activities include viewing of a product, rejecting of the product, purchase
of the product, initiation of checkout of the product and various other similar actions
performed by the users 102-106. In an embodiment of the present disclosure, the hybrid
recommendation system 114 is configured for recording behavior of the users 102-106 for
recommending the plurality of advertisements associated with the one or more products.
Further, the hybrid recommendation system 114 performs an item based recommendation, a
user based recommendation, demography based commendation and content based
recommendation (as described in detail in the detailed description of FIG. 2).
In an
embodiment of the present disclosure, the recommended plurality of advertisements is
displayed in a corresponding plurality of advertisement slots in the one or more publishers
110.
[0026]
In addition, the plurality of advertisements is provided by one or more advertisers
associated with the one or more publishers 110. The one or more advertisers purchase the
advertisement slots from the one or more publishers 110. In an embodiment of the present
disclosure, the one or more publishers 110 generate revenue based on one or more
compensation techniques. The one or more compensation techniques include but may not
be limited to pay per click, pay per view, cost per impression, cost per thousand impressions
and the like. In an embodiment of the present disclosure, the one or more publishers 110
and the one or more advertisers have a mutual contract for showing the plurality of
advertisements to the users 102-106.
[0027]
It may be noted that the users 102-106 are associated with the corresponding
communication devices 104-108 for accessing the one or more publishers 110 website,
however those skilled in the art would appreciate that there are more number of users
associated with more number of communication devices for accessing more number of
publishers.
For example, a user X, a user Y and a user Z are associated with a
communication device D1, a communication device D2 and a communication device D3.
[0028]
In an embodiment of the present disclosure, as illustrated in FIG. 1B, the hybrid
recommendation system 114 is a part of the one or more publishers 110. In an embodiment
of the present disclosure, the one or more publishers 110 include the hybrid recommendation
system 114. In addition, the hybrid recommendation system 114 is located on a backend of
each of the one or more publishers 114. In an embodiment of the present disclosure, the oneor more publishers 110 are registered on the hybrid recommendation system 114. In another
embodiment of the present disclosure, the one or more publishers 110 have an account on
the hybrid recommendation system 114. In an embodiment of the present disclosure, the
hybrid recommendation system 114 provides a web-based interface for the one or more
publishers 110. Moreover, the one or more publishers 110 utilize the web-based interface
for setting one or more preferences for targeting the one or more users for recommending
the plurality of advertisements.
[0029]
Gong further, in an embodiment of the present disclosure, the one or more
publishers 110 register on the hybrid recommendation system 114 by paying some pre-
defined amount of money in order to avail one or more services offered by the hybrid
recommendation system 114.
In an embodiment of the present disclosure, the hybrid
recommendation system 114 may accept multiple forms of payment to fund the account,
such as electronic transfer (e.g., automated clearing house (ACH) transfer or wire transfer)
from a designated bank account, credit card (e.g., Visa, MasterCard, Discover, American
Express), online wallet (e.g., PayPal, Amazon Payments and Google Checkout) and/or
mobile payment.
[0030]
In an embodiment of the present disclosure, the hybrid recommendation system 114
enables the one or more publishers 110 to download one or more comprehensive reports
associated with the one or more activities performed by the users 102-106 on the website.
In addition, the one or more publishers 110 are enabled to download reports associated with
the plurality of advertisements of the one or more products recommended to the users 102-
106. The reports help the one or more publishers 110 to gain valuable insight related to the
one or more user’s interest in the one or more products.
[0031]
FIG. 2 illustrates a block diagram 200 of a communication device 202, in
accordance with various embodiments of the present disclosure. The communication device
202 includes a processor 204, a control circuitry module 206, a storage module 208, an
input/output circuitry module 210 and a communication circuitry module 212. Further, the
processer 204 includes a collection module 204a, a clustering module 204b, a determination
module 204c, a gathering module 204d, a computation module 204e, an assigning module
204f, a recommendation engine 204g, a scoring module 204h, a decision module 204i, an
updation engine 204j and a database 204k. It may be noted that to explain the systemelements of FIG. 2, references will be made to the system elements of FIG. 1A and FIG.
1B. In an embodiment of the present disclosure, the processor 204 enables the working of
the hybrid recommendation system 114 for recommending the plurality of advertisements to
the users 102-106 viewing the one or more products on the corresponding publisher of the
one or more publishers 110.
In an embodiment of the present of the disclosure, the
communication device 202 enables the hosting of the hybrid recommendation system 114.
[0032]
Going further, the user 102 accesses a website of the publisher of the one or more
publishers 110 for accessing the content and viewing the one or more products. Moreover,
the collection module 204a in the processor 204 is configured to collect a first set of pre-
defined attributes associated with the user 102 of the one or more users viewing the one or
more products on the publisher of the one or more publishers 110. In addition, the first set
of pre-defined attributes include but may not be limited to an intent of the user 102, the one
or more products liked by the user 102, the one or more products disliked by the user 102,
age of the user 102, gender of the user 102, current location of the user 102, a type of device
utilized by the user 102, current contextual behavior of the user 102 and past behavior of the
user 102. In an embodiment of the present disclosure, the collection module 204a collects a
demographic information of the user 102 for classifying the user 102 into one or more
groups based on the first set of pre-defined attributes (as described below in detail in the
patent application). In an embodiment of the present disclosure, the collection is done for
determining the interests of the user 102 associated with a type of product of one or more
types of products. Moreover, the past behavior of the user 102 corresponds to the one or
more actions taken by the user 102 against the corresponding one or more products. In an
embodiment of the present disclosure, the past behavior illustrates affinity of the user 102 in
terms of the one or more products.
[0033]
Moreover, the first set of pre-defined attributes are collected based on dropping of a
cookie on the corresponding communication device 104 associated with the user 102. In an
embodiment of the present disclosure, the collection module 204a drops a cookie ID on the
communication device 106 when the user 102 accesses the website. The collection module
204a utilizes the cookie ID for extracting information associated with the user 102. In an
embodiment of the present disclosure, the user 102 is a frequent visitor on the publisher ofthe one or more publishers 114. In an embodiment of the present disclosure, the user 102 is
visiting the publisher website for the first time.
[0034]
For example, a user A accesses a publisher P1 (say Flipkart.com) on a browser X1
(say Google Chrome) on a communication device D1 (say a laptop) and a user B accesses a
publisher P2 (say Amazon.in) on a browser X2 (say Firefox) on a communication device D2
(say a smart phone). The collection module 204a collects the first set of pre-defined
attributes for the user A and the user B. The collection module 204a collects an age of the
user A and the user B (say, 22 for the user A and 25 for the user B), a gender of the user A
(say, male) and the user B (say, female), location of the user A (say, Delhi) and the user B
(say, Mumbai), products liked by the user A (say, shoes) and the user B (say, mobiles).
[0035]
Further, the clustering module 204b in the processor 204 is configured to classify the
user 102 of the one or more users in one or more clusters based on the first set of pre-
defined attributes. Each cluster of the one or more clusters corresponds to a group of users
of the one or more users having a similar liking associated with the product of the one or
more products. In an embodiment of the present disclosure, the cluster of the one or more
clusters may correspond to a plurality of products. In an embodiment of the present
disclosure, the clustering module 204b divides the one or more users based on the type of
the product liked by the one or more users and allotting a cluster of the one or more clusters
to each of the one or more users. In addition, each cluster of the one or more clusters
include users of the one or more users looking for a specific attribute in the product of the
one or more products. For example, a user X may be looking for blue colored sneakers of
the brand Puma, a user Y may be looking for a smart phone with 1GB Ram and 8 MP
camera and a user Z may be looking for a formal clothes of the brand Van Heusen and black
color.
[0036]
In an embodiment of the present disclosure, a number of clusters to be made is
performed using a hierarchical classification. In an embodiment of the present disclosure,
the number of clusters to be created is determined by using a dendogram technique.
Moreover, the one or more clusters are created by utilizing a K means cluster technique for
identifying the cluster of the one or more clusters corresponding to the product of the one or
more products currently viewed by the user 102.
In an embodiment of the present
disclosure, a new user (say user 106) of the one or more users with a new browsing historyis classified into the one or more clusters immediately. In an embodiment of the present
disclosure, profiling of each cluster of the one or more clusters is done based on means and
standard deviations of variables present in each of the cluster of the one or more clusters.
The profiling of the one or more clusters corresponds to assigning a type of the product
viewed by the user 102.
[0037]
In an embodiment of the present disclosure, the one or more clusters are already
present in the hybrid recommendation system 114 for classifying the user 102 into the one
or more clusters. Further, the one or more clusters are created based on a first set of
parameters. The first set of parameters include but may not be limited to demography of the
user 102, one or more details associated with the product of the one or more products
currently viewed by the user 102 and one or more attributes associated with the product of
the one or more products currently viewed by the user 102. The one or more attributes
differ for each type of the product viewed by the one or more users. Further, the one or
more attributes of the product are determined based on an identification of the cluster of the
one or more clusters (as described below in the patent application). In an embodiment of
the present disclosure, the user 102 may belong to a plurality of clusters based on the first
set of parameters. For example, a user K views mobile products and accessories and also
views one or more clothing products.
[0038]
Moreover, the classifying and the clustering of the user 102 are performed for the
identification of the cluster of the one or more clusters to which the user 102 belongs. In
addition, the identification of the cluster of the one or more clusters is based on calculation
of an average cluster centroid. Further, the calculation is done for determining the cluster of
the one or more clusters to which the user 102 is closest. In an embodiment of the present
disclosure, the average cluster centroid is calculated for a known user (the user 102)
searching and viewing the one or more products. Moreover, the one or more clusters are
created for identifying one or more parameters for the clustering.
The one or more
parameters aid the hybrid recommendation system 114 for finding out an attribute of the one
or more attributes associated with the product having a higher degree of importance for the
user 102.
[0039]
Going further, the one or more parameters are identified or determined using a
logistic regression technique or a discriminant analysis technique. In an embodiment of thepresent disclosure, the one or more parameters are identified for determining the attribute
which impacts a buying decision for the user 102. In addition, the hybrid recommendation
system 114 utilizes a stepwise discriminant analysis and a structure matrix for finding the
one or more parameters impacting the buying decision for the user 102. Moreover, the
clustering is done at pre-defined intervals of time.
In an embodiment of the present
disclosure, the set of parameters and the one or more parameters are analyzed at the pre-
defined intervals of time.
[0040]
In an embodiment of the present disclosure, the average cluster centroid for the user
102 is calculated by utilizing values of cluster centroid for a particular cluster corresponding
to the plurality of products belonging to the particular cluster viewed by the user 102 and
averaging the values of the cluster centroid. The cluster centroid corresponds to a middle of
the cluster. The calculated average cluster centroid value is used for determining the
product currently viewed by the user 102 by a degree of closeness of the average cluster
centroid value with the cluster centroid values for the particular cluster. Moreover, the
determining of the product is done by utilizing a product ID and a product name provided in
the cluster.
[0041]
Continuing the above stated example, the clustering module 204b classifies the user
A into a cluster C1 made up of Puma sneaker P, Nike sneaker N and Adidas sneaker AD and
classifies the user B into a cluster C2 made up of a mobile phone M1 (say Samsung Galaxy
Core 2 duos), a mobile phone M2 (say Samsung Galaxy Grand Prime) and a mobile phone
M3 (say Samsung Galaxy S5). The cluster centroid for the cluster C1 for the puma sneaker
P is 6.17544, the Nike sneaker N is 10.82456 and the Adidas sneaker AD is 11.82456.
Moreover, the cluster centroid for the cluster C2 for the mobile phone M1 is 6.17544, for the
mobile phone M2 is 10.82546 and for the mobile phone M3 is 0.82546. The average cluster
centroid for the cluster C1 for the user A is 9.61 and for the cluster C2 for the user B is 6.27.
The cluster module 204b determines that the user A is closest to the cluster for the Nike
sneaker N and determines that the user B is closest to the cluster for the mobile phone M1
(Samsung Galaxy Core 2 duos). The clustering module 204b determines that the user A is
viewing the Nike sneaker N based on product ID and name given in the cluster C1 for the
Nike sneaker N and determines that the user B is viewing the mobile phone M1 (SamsungGalaxy Core 2 duos) based on product ID and name given in the cluster C2 for the mobile
phone M1 (Samsung Galaxy Core 2 duos).
[0042]
Going further, the determination module 204c in the processor 204 is configured for
determining one or more characteristics associated with the product of the one or more
products currently viewed by the user 102 of the one or more users. The one or more
characteristics correspond to the one or more attributes associated with the product of the
one or more products. In addition, the one or more characteristics are different for each type
of the product of the one or more products.
The determining of the one or more
characteristics associated with the product of the one or more products is based on the
identification of the cluster of the one or more clusters by the calculation of the average
cluster centroid. The one or more characteristics associated with the product currently
viewed by the user 102 are determined by the identification of the product through the
product ID and the product name provided in the cluster corresponding to the product
determined by the average cluster centroid value (as exemplary stated above in the patent
application).
[0043]
In an embodiment of the present disclosure, the determination module 204c
determines a limited number of characteristics of the corresponding product. For example, a
user J viewing a smartphone (say Apple iPhone 5S) on a website. The determination
module 204c determines model name and model ID of the smart phone and determines
some characteristics (say, brand, color and price) of the one or more characteristics
associated with the smartphone. In an embodiment of the present disclosure, in such a case,
the determination module 204c utilizes a scrapped data and look alike modeling for
determining the one or more characteristics and information related to the product searched
by the user 102. In an embodiment of the present disclosure, this approach is utilized by the
publisher of the one or more publishers 110 not associated with e-commerce background.
For example, the user J accesses a website (say groupon.com) for accessing coupons for
smartphones. The user J views a coupon for an Apple iPhone 5S device. The determination
module 204c determines that the user J is looking for smartphones of the brand Apple and
the phone Apple iPhone 5S.
[0044]
In addition, the gathering module 204d in the processor 204 is configured to gather
data associated with other one or more products similar to the product currently viewed bythe user 102 based on the determination of the one or more characteristics and the
identification of the cluster of the one or more clusters. Moreover, the other one or more
products have similar characteristics to the one or more characteristics associated with the
product of the one or more products currently viewed by the user 102. In an embodiment of
the present disclosure, the other one or more products correspond to a similar product
category associated with the product currently viewed by the user 102. Moreover, the other
one or more products are determined using the one or more characteristics determined for
the product of the one or more products currently viewed by the user 102 and finding similar
products (the other one or more products) having the one or more characteristics.
[0045]
In an embodiment of the present disclosure, accuracy of the gathered data associated
with the other one or more products is based on a level of the one or more characteristics
determined for the product of the one or more products currently viewed by the user 102.
The level of the one or more characteristics corresponds to a number of significant
characteristics which will help in finding out the other one or more products similar to the
product currently viewed by the user 102. Further, the other one or more products belong to
the identified cluster of the one or more clusters corresponding to the product of the one or
more products currently viewed by the user 102.
[0046]
In an embodiment of the present disclosure, the other one or more products are
identified in a separate cluster having the similar characteristics of the one or more
characteristics. Moreover, the other one or more products are identified by using a cosine
similarity technique.
In an embodiment of the present disclosure, the hybrid
recommendation system 114 utilizes the cosine similarity technique for calculating a value
of similarity between the product of the one or more products currently viewed by the user
102 with each of the other one or more products identified.
[0047]
Extending the above stated example, the user B is currently viewing at the mobile
phone M1 (Samsung Galaxy Core 2 duos) as determined by the clustering module 204c.
The determination module 204c determines the one or more characteristics associated with
the mobile phone M1 (say, model id is SM-G355HZWDINU, mobile id is 2, brand is
Samsung, color is white, operating system is Android and the like). The gathering module
204d utilizes the one or more characteristics of the mobile phone M1 to find other mobile
phones (say, A106, A501CG- Black, A501CG- Blue, A501CG- White, A501CG- Red, X2-green, X2-Black, X2- White, X2- Yellow, XT1022, AQ4501- Black, AQ4501- White, X5-
Black, X5- White and A104-Grey).
[0048]
In an embodiment of the present disclosure, the other one or more products are
identified based on a propensity to buy. The propensity to buy is pre-calculated based on
interactions of other one or more users with the other one or more products. In addition, the
other one or more products having the highest propensity to buy are selected (as described
below in the patent application).
[0049]
Going further, the computational module 204e in the processor 204 is configured to
compute a weighted average propensity to buy for the user 102 of the one or more users.
The weighted average propensity to buy corresponds to a probability of the user 102 buying
the product of the one or more products currently viewed by the user 102. In addition, the
weighted average propensity to buy is computed for the product of the one or more products
currently viewed by the user 102 of the one or more users and the other one or more
products similar to the product currently viewed by the user 102. Moreover, the weighted
average propensity to buy is computed based on a second set of pre-defined attributes and a
pre-defined criterion.
Further, the pre-defined criterion is based on analyzing a past
behavior of a set of users of the one or more users who have bought the product of the one
or more products and the other one or more products and taken one or more actions
corresponding to the product of the one or more products and the other one or more
products.
[0050]
Furthermore, the second set of pre-defined attributes corresponds to one or more
types of events taken place during interaction of the set of users with the product of the one
or more products and the other one or more products. The one or more types of events
include but may not be limited to a duration of view by the user 102 and the set of users for
the product of the one or more products and the other one or more products, search
performed by the user 102 and the set of users for searching the product of the one or more
products and the other one or more products, add to cart event, add to wish list event, a
purchase event, checkout initiated event, product view and product reject event.
[0051]
In an embodiment of the present disclosure, the weighted average propensity to buy
is calculated by utilizing a number of users corresponding to each of the one or more types
of events for each of the other one or more products and the product currently viewed by theuser 102. In addition, the number of users is multiplied by a propensity to buy value
assigned for each of the one or more types of events (as described later in the patent
application) for calculating the weighted average propensity to buy. In an embodiment of
the present disclosure, the weighted average propensity to buy is calculated for a fixed
interval of time.
[0052]
In an embodiment of the present disclosure, the weighted average propensity to buy
is pre-calculated for the other one or more products. In an embodiment of the present
disclosure, the weighted average propensity to buy may change for the product of the one or
more products currently viewed by the user 102.
In an embodiment of the present
disclosure, the weighted average propensity to buy for the other one or more products is
fetched. In another embodiment of the present disclosure, the other one or more products
with the highest weighted average propensity to buy are fetched.
[0053]
Continuing the above stated example, the computational module 204e calculates the
weighted average propensity to buy for the mobile phone M1 (Samsung Galaxy Core 2
Duos) and the other mobile phone (say, A106, A501CG- Black, A501CG- Blue, A501CG-
White, A501CG- Red, X2-green, X2-Black, X2- White, X2- Yellow, XT1022, AQ4501-
Black, AQ4501- White, X5- Black, X5- White, A104-Grey and A104-Grey). The weighted
average propensity to buy for the mobile phone M1 (Samsung Galaxy Core 2 Duos) is
calculated by adding number of users for each of the one or more types of events (say, 10 for
purchased event, 11 for checkout initiated event, 12 for add to cart event, 4 for add to wish
list event, 28 for product view event, 2 for product search event and 39 for product reject
event). The number of users is 106. Further, the weighted average propensity to buy is
calculated by dividing an output of addition of product of the number of users for each of
the one or more types of events with the propensity to buy value assigned for each of the
one or more types of events (say 501) with the total number of users for the mobile phone
M1 (106). The weighted average propensity to buy is 4.726. Similarly, the computational
module 204e calculates the weighted average propensity to buy for the other mobile phone
(say, A106, A501CG- Black, A501CG- Blue, A501CG- White, A501CG- Red, X2-green,
X2-Black, X2- White, X2- Yellow, XT1022, AQ4501- Black, AQ4501- White, X5- Black,
X5- White and A104-Grey).[0054]
Going further, the assigning module 204f in the processor 204 is configured for
assigning a value corresponding to each of the one or more types of events for the user and
the set of users. The value is assigned for determining the propensity to buy for the product
of the one or more products currently viewed by the user 102 and the other one or more
products. Moreover, the assigned value is highest for the purchase event and the value is
lowest for the product reject event. In an embodiment of the present disclosure, the value is
utilized for the calculation of the weighted average propensity to buy for the product of the
one or more products currently viewed by the user 102 and the other one or more products.
In an embodiment of the present disclosure, the assigning module 204f assigns a value of 10
for the purchase event, a value of 9 for the checkout initiated event, a value of 8 for the add
to cart event, a value of 7 for the add to wish list event, a value of 6 for the product view
event, a value of 5 for the product search event and a value of 0 for the product reject event.
[0055]
Moreover, the scoring module 204g in the processor 204 is configured to calculate a
recommendation score for the product of the one or more products currently viewed by the
user 102 of the one or more users and the other one or more products. In addition, the
recommendation score is calculated based on multiplication of the similarity of the product
currently viewed by the user 102 with the product currently viewed by the user 102 and the
other one or more products and the corresponding weighted average propensity to buy for
the product currently viewed by the user 102 and the other one or more products.
[0056]
Extending the above stated example, the propensity to buy for the mobile phone M1
is 0.5. The similarity of the mobile phone M1 with the mobile phone M1 is 1. The scoring
module 204g calculates the recommendation score for the mobile phone M1 (1*0.5=0.5).
Similarly, the scoring module 204g calculates the recommendation score for the A106
(0.999*0.75=0.749), A501CG- Black (0.999*1=0.999), A501CG- Blue (0.999*1=0.999),
A501CG-
White
(0.999*0.5=0.500),
A501CG-
Red
(0.999*0=0.000),
X2-green
(1.000*1=1.000), X2-Black (1.000*1=1.000), X2- White (1.000*1=1.000), X2- Yellow
(1.000*0=0.000), XT1022 (0.999*1=0.999), AQ4501- Black (1.000*1=1.000), AQ4501-
White (1.000*1=1.000), X5- Black (0.999*1=0.999), X5- White (0.999*1=0.999) and
A104-Grey (0.999*1=0.999).
[0057]
In an embodiment of the present disclosure, the hybrid recommendation system 114
sorts the product and the other one or more products based on the recommendation score ina descending order. The sorting is done based on the weighted average propensity to buy
and the calculated recommendation score. In an embodiment of the present disclosure, the
hybrid recommendation system 114 selects a pre-defined number of products from the
sorted list of the product and the other one or more products. In an example, the hybrid
recommendation system 114 selects a top 10 products from the sorted list for
recommendation to the user 102.
[0058]
Going further, the recommendation engine 204h in the processor 204 is configured
to recommend the plurality of advertisements associated with the product of the one or more
products currently viewed by the user 102 and the other one or more products having the
highest weighted average propensity to buy. In an embodiment of the present disclosure, the
recommendation engine 204h recommends the plurality of advertisements based on the
highest recommendation score.
The plurality of advertisements may be a banner
advertisement, a textual advertisement, html advertisement and the like. In an embodiment
of the present disclosure, the plurality of advertisements is displayed in the corresponding
plurality of advertisement slots in the one or more publishers 110. Moreover, the plurality
of advertisements is provided by the one or more advertisers in real time.
[0059]
In an example, the recommendation engine 204h recommends the top 10 products
through the plurality of advertisements from the sorted list of the products and the other one
or more products. Further, the recommendation of the plurality of advertisements associated
with the product and the other one or more products having the highest weighted average
propensity to buy is based on seller rating, a price of the product of the one or more products
and the other one or more products recommended to the user 102, click through rate and a
seller bid for a particular ad format and publisher combination.
[0060]
In an embodiment of the present disclosure, each of the product and the other one or
more products are recommended through a corresponding single advertisement. In an
embodiment of the present disclosure, the product currently viewed by the user 102 on an e-
commerce publisher of the one or more publishers 110 is not recommended to the user 102
in case of no third party product listing ad is allowed by the e-commerce publisher. In an
embodiment of the present disclosure, the recommendation of the product currently viewed
by the user 102 is allowed if the e-commerce publisher allows the third party product listingad and if the recommendation score recommends the product from other advertiser of the
one or more advertisers.
[0061]
In an embodiment of the present disclosure, the recommendation engine 204h does
not recommend products which are already bought by the user 102. In an embodiment of
the present disclosure, the recommendation engine 204h recommends recently viewed
products to the user 102.
In an embodiment of the present disclosure, the hybrid
recommendation system 114 performs the recommendation of the recently viewed products
by two methods. In an embodiment of the present disclosure, first method of the two
methods is by finding the similarity between the product of the one or more products
currently viewed by the user 102 with the recently viewed products and calculating a new
propensity to buy for each of the recently viewed products by utilizing data associated with
the other one or more users. In another embodiment of the present disclosure, second
method of the two methods works by recommending the recently viewed products
irrespective of the similarity with the product currently viewed by the user 102.
[0062]
Extending the above stated example, the recommendation engine 204h recommends
the other mobile phones (X2-green, X2-Black, X2-White, AQ4501-Black and AQ4501-
White) having the highest weighted average propensity to buy (1.000) for the user B.
[0063]
Further, the decision module 204i in the processor 204 is configured for deciding an
advertiser of the one or more advertisers whose advertisement of the plurality of
advertisements corresponding to the recommended product of the one or more products and
the other one or more products will be displayed to the user 102 of the one or more users.
Furthermore, the decision is taken based on a condition specified by the publisher of the one
or more publishers 110. The condition specified by the publisher of the one or more
publishers 110 corresponds to a choice of the publisher for allowing one or more third party
product listing advertisements. In an embodiment of the present disclosure, the decision
module 204i takes the decision based on when the publisher of the one or more publishers
110 does not allow the one or more third party product listing advertisements. In another
embodiment of the present disclosure, the decision module 204i takes the decision based on
when the publisher of the one or more publishers 110 allows the one or more third party
product listing advertisements.[0064]
Going further, in an embodiment, when the publisher of the one or more publishers
110 does not allow the one or more third party product listing advertisements on the website,
the decision module 204i checks availability of the recommended product of the one or
more product and the other one or more products with the publisher and recommends the
available products to the user 102 through the corresponding advertisements of the plurality
of advertisements. In an embodiment of the present disclosure, the decision module 204i
performs the decision based on the availability of the product and the other one or more
products with a single seller of the publisher of the one or more publishers 110.
[0065]
In an embodiment of the present disclosure, in this case, the decision module 204i
takes the decision based on the ad format of a plurality of ad formats having a highest value
of product of click through rate (CTR) and click rate (CR). In an embodiment of the present
disclosure, the decision module 204i analyzes the click through rate (CTR) and the click rate
(CR) for each of the plurality of ad formats.
Further, the recommended product is
recommended in the ad format having the highest value of the product of click through rate
(CTR) and click rate (CR). In an embodiment of the present disclosure, the decision module
204i displays the recommended product in the ad format pre-selected by the publisher of the
one or more publishers 110.
[0066]
In an embodiment of the present disclosure, the decision module 204i performs the
decision based on the availability of the recommended product and the other one or more
products with a plurality of sellers of the publisher of the one or more publishers 110.
Further, in this case, the decision module 204i takes the decision based on deciding a seller
of the plurality of sellers whose product will be displayed. Moreover, the decision is taken
based on a seller rating of each of the plurality of sellers, a delivery time for each of the
plurality of sellers, a price of the product for each of the plurality of sellers, a margin to the
advertiser of the one or more advertisers for each of the plurality of sellers, an advertisement
budget for each of the plurality of sellers and a bidding price for each of the plurality of
sellers. In addition, the decision module 204i calculates a final score for each of the
plurality of sellers. Further, the decision module 204i displays the advertisement of the
recommended product corresponding to the seller of the plurality of sellers having the
highest final score.[0067]
Furthermore, in an embodiment, when the publisher of the one or more publishers
110 allows the one or more third party product listing advertisements on the website, the
decision module 204i takes the decision based on whether the recommended product is
available with a single advertiser of the one or more advertisers and whether the
recommended product is available with a plurality of advertisers. In an embodiment of the
present disclosure, the decision module 204i recommends the product available with the
single advertiser after deciding of the ad format of a plurality of ad formats having the
highest value of the product of the click through rate (CTR) and the click rate (CR).
[0068]
In another embodiment of the present disclosure, the decision module 204i takes the
decision for the recommended product available with the plurality of advertisers based on a
plurality of attributes. The plurality of attributes include but may not be limited to a sale
price of the recommended product for each of the plurality of advertisers, a sale price
inverse of the recommended product for each of the plurality of advertisers, one or more
creative type associated with each of the plurality of advertisers, product of the click through
rate (CTR) and the click rate (CR) for each of the one or more creative type for each of the
plurality of advertisers, an advertiser bid value for each of the plurality of advertisers and a
percentage of margin received by the publisher of the one or more publishers 110.
[0069]
Further, a final recommendation score is calculated based on the plurality of
attributes. Moreover, the final recommendation score is calculated based on multiplication
of the sale price inverse, the product of the click through rate (CTR) and the click rate (CR),
the advertiser bid value and the percentage of margin received by the publisher. In addition,
the decision module 204i recommends the product for the advertiser of the one or more
advertiser having the highest final recommendation score.
[0070]
In an embodiment of the present disclosure, when a new user of the one or more user
accesses the publisher of the one or more publishers 110, the hybrid recommendation system
114 utilizes demography of the new user, details of a communication device utilized for
accessing the publisher and one or more attributes of the product currently viewed by the
new user and the new user is classified using into the one or more clusters.
In an
embodiment of the present disclosure, the one or more clusters are created based on the
demography if the one or more parameters cannot be identified for the clustering.[0071]
In an embodiment of the present disclosure, when a new product of the one or more
products is introduced by the publisher of the one or more publishers 110 for which no user
behavior has been captured by the hybrid recommendation system 114, the new product is
classified into the one or more clusters based on the one or more attributes of the new
product found using the discriminant analysis method. Moreover, the propensity to buy for
the new product is calculated based on the one or more attributes in the identified cluster. In
addition, a multinomial linear regression method is utilized for calculating the propensity to
buy for the new product. In an embodiment of the present disclosure, the new product is re-
classified based on capturing of the user behavior for the new product and the one or more
attributes of the new product.
[0072]
Going further, the updation engine 204j in the processor 204 is configured to update
the weighted average propensity to buy, the one or more clusters, the first set of pre-defined
attributes, the first set of parameters, the second set of pre-defined attributes and the
recommendation score. The updation is performed at pre-defined continuous intervals of
time. The updation is done for refining of recommendation algorithm. In addition, the
database 204k in the processor 204 is configured for storing the weighted average
propensity to buy, the one or more clusters, the first set of pre-defined attributes, the first set
of parameters, the second set of pre-defined attributes and the recommendation score.
[0073]
It may be noted that in FIG. 2, various modules of the hybrid recommendation
system 114 are shown that illustrates the working of the hybrid recommendation system
114; however those skilled in the art would appreciate that the hybrid recommendation
system 114 may have more number of modules that could illustrate overall functioning of
the hybrid recommendation system 114.
[0074]
Going further, the communication device 202 includes any suitable type of portable
electronic device. Examples of the communication device 202 include but may not be
limited to a personal e-mail device (e.g., a Blackberry.TM. made available by Research in
Motion of Waterloo, Ontario), a personal data assistant ("PDA"), a cellular telephone, a
Smartphone, the laptop computer, and the tablet computer. In another embodiment of the
present disclosure, the communication device 202 can be a desktop computer.
[0075]
From the perspective of this disclosure, the control circuitry module 206 includes
any processing circuitry or processor operative to control the operations and performance ofthe communication device 202. For example, the control circuitry module 206 may be used
to run operating system applications, firmware applications, media playback applications,
media editing applications, or any other application.
[0076]
In an embodiment, the control circuitry module 206 drives a display and process
inputs received from the user interface.
[0077]
From the perspective of this disclosure, the storage module 208 includes one or
more storage mediums including a hard-drive, solid state drive, flash memory, permanent
memory such as ROM, any other suitable type of storage component, or any combination
thereof. The storage module 208 may store, for example, media data (e.g., music and video
files), application data (e.g., for implementing functions on the communication device 202).
[0078]
From the perspective of this disclosure, the I/O circuitry module 210 may be
operative to convert (and encode/decode, if necessary) analog signals and other signals into
digital data. In an embodiment, the I/O circuitry module 210 may also convert the digital
data into any other type of signal and vice-versa. For example, the I/O circuitry module 210
may receive and convert physical contact inputs (e.g., from a multi-touch screen), physical
movements (e.g., from a mouse or sensor), analog audio signals (e.g., from a microphone),
or any other input. The digital data may be provided to and received from the control
circuitry module 206, the storage module 208 or any other component of the communication
device 202.
[0079]
It may be noted that the I/O circuitry module 210 is illustrated in FIG. 2 as a single
component of the communication device 202; however those skilled in the art would
appreciate that several instances of the I/O circuitry module 210 may be included in the
communication device 202.
[0080]
The communication device 202 may include any suitable interface or component for
allowing the user 102 to provide inputs to the I/O circuitry module 210.
The
communication device 202 may include any suitable input mechanism. Examples of the
input mechanism include but may not be limited to a button, keypad, dial, a click wheel, and
a touch screen. In an embodiment, the communication device 202 may include a capacitive
sensing mechanism, or a multi-touch capacitive sensing mechanism.
[0081]
In an embodiment, the communication device 202 may include specialized output
circuitry associated with output devices such as, for example, one or more audio outputs.The audio output may include one or more speakers built into the communication device
202, or an audio component that may be remotely coupled to the communication device
202.
[0082]
The one or more speakers can be mono speakers, stereo speakers, or a combination
of both. The audio component can be a headset, headphones or ear buds that may be
coupled to the communication device 202 with a wire or wirelessly.
[0083]
In an embodiment, the I/O circuitry module 210 may include display circuitry for
providing a display visible to the user 102. For example, the display circuitry may include a
screen (e.g., an LCD screen) that is incorporated in the communication device 202.
[0084]
The display circuitry may include a movable display or a projecting system for
providing a display of content on a surface remote from the communication device 202
(e.g., a video projector).
In an embodiment, the display circuitry may include a
coder/decoder to convert digital media data into the analog signals. For example, the
display circuitry may include video Codecs, audio Codecs, or any other suitable type of
Codec.
[0085]
The display circuitry may include display driver circuitry, circuitry for driving
display drivers or both. The display circuitry may be operative to display content. The
display content can include media playback information, application screens for applications
implemented on the electronic device, information regarding ongoing communications
operations, information regarding incoming communications requests, or device operation
screens under the direction of the control circuitry module 206. Alternatively, the display
circuitry may be operative to provide instructions to a remote display.
[0086]
In addition, the communication device 202 includes the communication circuitry
module 212.
The communication circuitry module 212 may include any suitable
communication circuitry operative to connect to a communication network and to transmit
communications (e.g., voice or data) from the communication device 202 to other devices
within the communications network. The communication circuitry module 212 may be
operative to interface with the communication network using any suitable communication
protocol. Examples of the communication protocol include but may not be limited to Wi-Fi,
Bluetooth RTM, radio frequency systems, infrared, LTE, GSM, GSM plus EDGE, CDMA,
and quadband.[0087]
In an embodiment, the communication circuitry module 212 may be operative to
create a communications network using any suitable communications protocol.
For
example, the communication circuitry module 212 may create a short-range communication
network using a short-range communications protocol to connect to other devices. For
example, the communication circuitry module 212 may be operative to create a local
communication network using the Bluetooth, RTM protocol to couple the communication
device 202 with a Bluetooth, RTM headset.
[0088]
It may be noted that the computing device is shown to have only one communication
operation; however, those skilled in the art would appreciate that the communication device
202 may include one more instances of the communication circuitry module 212 for
simultaneously
performing
several
communication
operations
using
different
communication networks. For example, the communication device 202 may include a first
instance of the communication circuitry module 212 for communicating over a cellular
network, and a second instance of the communication circuitry module 212 for
communicating over Wi-Fi or using Bluetooth RTM.
[0089]
In an embodiment of the present disclosure, the same instance of the communication
circuitry module 212 may be operative to provide for communications over several
communication networks.
In an embodiment, the communication device 202 may be
coupled to a host device for data transfers, syncing the communication device 202, software
or firmware updates, providing performance information to a remote source (e.g., providing
riding characteristics to a remote server) or performing any other suitable operation that may
require the communication device 202 to be coupled to the host device. Several computing
devices may be coupled to a single host device using the host device as a server.
Alternatively or additionally, the communication device 202 may be coupled to the several
host devices (e.g., for each of the plurality of the host devices to serve as a backup for data
stored in the communication device 202).
[0090]
FIG. 3 illustrates an example snapshot 300 showing classification of the one or
more users into one or more clusters, in accordance with various embodiments of the
present disclosure. Moreover, the snapshot shows the classification of the one or more users
(a user 1, a user 2, a user 3 and a user 4) into the one or more clusters (a cluster 1 and a
cluster 2) based on a model id and a model name. The one or more users view one or moremobile phones. Further, the user 1 belongs to the cluster 1 and the cluster2. Similarly, the
user 2, the user 3 and the user 4 belongs to the cluster 1 and the cluster 2. In addition, the
user 1 is associated with viewing of three mobile devices (Samsung Galaxy Core 2 Duos,
Samsung Galaxy Grand Prime and Samsung Galaxy S5).
[0091]
Moreover, the average cluster centroid for each of the mobile device for the
corresponding cluster is provided. In addition, the average cluster centroid for the user 1
associate with the cluster 1 is determined by taking the average of the cluster centroid for
the cluster 1 corresponding to the user 1. In an example, as shown in FIG. 3, the average
cluster centroid for the user 1 associated with the cluster 1 is 6.27. The average cluster
centroid value determines that the user 1 is viewing the Samsung Galaxy Core 2 Duos
mobile device with the model id SM-G355HZWDINU.
[0092]
FIG. 4 illustrates an example snapshot 400 for showing the calculation of the
weighted average propensity to buy, in accordance with various embodiments of the present
disclosure. As shown in the FIG. 4, the weighted average propensity to buy is calculated
for the product with the model id SM-G355HZWDINU based on the one or more types of
events for the set of users who have interacted with the Samsung Galaxy Core 2 Duos
mobile device. The number of users is provided for each of the one or more types of events.
Moreover, the propensity to buy value is provided for each of the one or more types of
events. In addition, the product of the number of users and the propensity to buy is provided
for each of the one or more types of events. Further, the weighted average propensity to buy
is calculated by dividing the product of the number of users and the propensity to buy (501)
with the number of users (106). The weighted average propensity to buy is 4.72641509 for
the Samsung Galaxy Core 2 Duos mobile device.
[0093]
FIG. 5 illustrates an example snapshot 500 for calculating the recommendation
score, in accordance with various embodiments of the present disclosure. Moreover, the
recommendation score is calculated for each of the other one or more products and the
products of the one or more products currently viewed by the user 102 (Samsung Galaxy
Core 2 Duos). In addition, the recommendation score is calculated by multiplying the
similarity of the product of the one or more products currently viewed by the user 102
(Samsung Galaxy Core 2 Duos) with the other one or more products (820, A106, A501CG-
Black, A501CG- Blue, A501CG- White, A501CG- Red, X2-green, X2-Black, X2- White,X2- Yellow, XT1022, AQ4501- Black, AQ4501- White, X5- Black, X5- White and A104-
Grey) with the propensity to buy for each of the corresponding other or more products and
the product of the one or more products currently viewed by the user 102.
The
recommendation score for each of the other one or more products and the product currently
viewed is provided.
[0094]
FIG. 6A, FIG. 6B and FIG. 6C illustrates an example snapshot 600 for taking a
decision associated with the advertiser of the one or more advertisers for the recommending
of the plurality of advertisements, in accordance with various embodiments of the present
disclosure. FIG. 6A illustrates a snapshot 600 showing the decision associated with the
displaying of the recommended product in the ad format having the highest value of the
product of the click through rate and the click rate for the publisher (Snapdeal) not allowing
the one or more third party product listing advertisement and having the single seller
available for the recommended product. As shown in the FIG. 6A, the recommended
product is displayed in the tinder format having the highest value of the product of the click
through rate and the click rate of 0.35.
[0095]
FIG. 6B illustrates a snapshot 600 showing the decision associated with the
displaying of the recommended product for the publisher (Snapdeal) not allowing the one or
more third party product listing advertisement and having the plurality of sellers available
for the recommended product. As shown in the FIG. 6B, the plurality of sellers include
seller1, seller 2, seller 3, seller 4 and seller 5. The final score for each of the plurality of
sellers is calculated based on the seller rating, the delivery time, the price of the
recommended product, the margin to advertiser and the advertisement budget. In addition, a
seller (seller 4) of the plurality of sellers having the highest final score (59708.97) is
selected for displaying the advertisement associated with the recommended product.
[0096]
FIG. 6C illustrates a snapshot 600 for showing the decision taken for the publisher
of the one or more publishers 110 allowing the one or more third party product listing
advertisement and having the plurality of advertisers for providing the advertisements for
the corresponding recommended product (X2-green). As shown in the FIG. 6C, the final
recommendation score is calculated for the recommended product (X2-green) based on the
multiplication of the sale inverse price P, the value A of the product of the click through rate
(CTR) and the click rate (CR), the advertiser bid value B and the percentage M that thepublisher gets. The advertisement of the advertiser Amazon in the Ad format 3 is selected
based on the final recommendation score of the advertiser Amazon in the Ad format 3 is the
highest final recommendation score.

Claims:
What is claimed is:
1. A hybrid recommendation system for recommending a plurality of advertisements to a user of
one or more users viewing one or more products on a publisher of one or more publishers, the
hybrid recommendation system comprising:
a collection module in a processor, the collection module being configured to collect a
first set of pre-defined attributes associated with the user of the one or more users viewing the
one or more products on the publisher of the one or more publishers;
a clustering module in the processor, the clustering module being configured to classify
the user of the one or more users in one or more clusters based on the first set of pre-defined at-
tributes, wherein the one or more clusters being created based on a first set of parameters,
wherein the classifying and clustering being done for identifying a cluster of the one or more
clusters to which the user of the one or more users belongs, wherein the identifying being based
on calculation of an average cluster centroid value for the user of the one or more users, wherein
the one or more clusters being created for identifying one or more parameters for the clustering
and wherein the clustering being done at a pre-defined intervals of time;
a determination module in the processor, the determination module being configured to
determine one or more characteristics associated with a product of the one or more products cur-
rently viewed by the user of the one or more users, wherein the determining of the one or more
characteristics being based on the identification of the cluster of the one or more clusters;
a gathering module in the processor, the gathering module being configured to gather
data associated with other one or more products similar to the product currently viewed by the
user based on the determination of the one or more characteristics and the identification of the
cluster of the one or more clusters, wherein the other one or more products depict similar charac-
teristics to the one or more characteristics associated with the product of the one or more prod -
ucts currently viewed by the user and wherein the other one or more products belong to an iden-
tified cluster of the one or more clusters corresponding to the product of the one or more prod-
ucts currently viewed by the user;
a computational module in the processor, the computational module being configured to
compute a weighted average propensity to buy for the user of the one or more users, wherein the
weighted average propensity to buy being computed for the product of the one or more products
currently viewed by the user of the one or more users and the other one or more products similarto the product currently viewed by the user, wherein the weighted average propensity to buy be-
ing computed based on a second set of pre-defined attributes and a pre-defined criterion, wherein
the pre-defined criterion being based on analyzing a past behavior of a set of users of the one or
more users who have bought the product of the one or more products and the other one or more
products and taken one or more actions corresponding to the product of the one or more products
and the other one or more products; and
a recommendation engine in the processor, the recommendation engine being configured
to recommend the plurality of advertisements associated with the product of the one or more
products currently viewed by the user and the other one or more products having a highest
weighted average propensity to buy.
2. The hybrid recommendation system as recited in claim 1, further comprising a scoring module in
the processor, the scoring module being configured to calculate a recommendation score for the
product of the one or more products currently viewed by the user of the one or more users and
the other one or more products, wherein the recommendation score being calculated based on
multiplication of a similarity of the product currently viewed by the user with the product cur-
rently viewed by the user and the other one or more products and the corresponding weighted av-
erage propensity to buy for the product currently viewed by the user and the other one or more
products.
3. The hybrid recommendation system as recited in claim 2, wherein the recommendation being
based on at least one of a seller rating, a price of the product of the one or more products and the
other one or more products recommended to the user, click through rate and a seller bid for a par-
ticular ad format and publisher combination.
4. The hybrid recommendation system as recited in claim 1, wherein the first set of pre-defined at-
tributes comprises at least one of an intent of the user, the one or more products liked by the user,
the one or more products disliked by the user, age of the user, gender of the user, current location
of the user, a type of device utilized by the user, current contextual behavior of the user and past
behavior of the user.
5. The hybrid recommendation system as recited in claim 1, wherein the first set of parameters
comprises at least one of a demography of the user, one or more details associated with the prod -
uct of the one or more products and one or more attributes associated with the product of the one
or more products.
6. The hybrid recommendation system as recited in claim 1, wherein the second set of pre-defined
attributes corresponds to one or more types of events, wherein the one or more types of eventscomprises at least one of a duration of view by the user and the set of users for the product of the
one or more products and the other one or more products, search performed by the user and the
set of users for searching the product of the one or more products and the other one or more
products, add to cart event, add to wish list event, a purchase event, checkout initiated event,
product view and product reject event.
7. The hybrid recommendation system as recited in claim 6, further comprising an assigning mod-
ule in the processor, the assigning module being configured for assigning a value corresponding
to each of the one or more types of events for the user and the set of users, wherein the value be -
ing assigned for determining a propensity to buy for the product of the one or more products cur-
rently viewed by the user and the other one or more products and wherein the value being high-
est for the purchase event and the value being lowest for the product reject event.
8. The hybrid recommendation system as recited in claim 1, wherein the one or more characteristics
associated with the product of the one or more products comprises at least one of a type of the
product viewed, a category of the product viewed, a name of the product viewed, an id of the
product viewed, a brand name of the product viewed and one or more specifications of the prod-
uct viewed by the user.
9. The hybrid recommendation system as recited in claim 1, further comprising an updation engine
in the processor, the updation engine being configured to update the weighted average propensity
to buy, the one or more clusters, the first set of pre-defined attributes, the first set of parameters,
the second set of pre-defined attributes and the recommendation score, wherein the updation be-
ing performed at a pre-defined continuous intervals of time and wherein the updation being done
for refining of recommendation algorithm.
10. The hybrid recommendation system as recited in claim 1, further comprising a decision module
in the processor, the decision module being configured for deciding an advertiser of one or more advertisers whose advertisement of the plurality of advertisements corresponding to the recommended product of the one or more products and the other one or more products will be dis-
played to the user of the one or more users, wherein the decision being taken based on a condition specified by the publisher of the one or more publishers.