A SYSTEM FOR PREDICTING CREDIT WORTHINESS OF A FINANCIAL ENTITY BASED ON AN ENHANCED ADAPTIVE DATA MODEL AND HISTORICAL DATA
ANALYTICS
Reference of the Provisional Application:
This application takes reference from the provisional application 3342/MUM/2014, Docket Number 21335, submitted on 20th October 2014 in Mumbai.
BACKGROUND OF THE INVENTION
1. Technical Field
The present invention relates generally to credit evaluation and rating models, and more specifically relates to a system for employing analytics modeling to evolve credit worthiness of a financial entity.
2. Related Art
Currently, there are several methods and tools developed from various financial data elements
for calculating and evaluating credit worthiness and credit ratings. These ratings are used for
various purposes of granting loans, investments, stock market and borrowing and lending
purposes.
The methodologies and tools available are generic in nature and employ various tools and
techniques including machine learning and statistics.

SUMMARY OF THE INVENTION
.4
A system for predicting credit worthiness of a financial entity based on an adaptive enhanced data model using a set of attributes including at least one of: evaluation of depreciating assets; indirect social or indirect professional connection; interdependent multivariate analysis of various financial parameters; evaluation of business need for at least one asset. The data model further including parameterized customer background; business model parameters; asset specific parameters; quantified socio-geographical parameters and any combination thereof. The data model further includes financial performance metrics including but not limited to macro-financial performance indicators, micro-financial performance indicators and temporal financial performance indicators. The data model further may include projected performance parameters for the financial entity.
As per one embodiment, the relevant data is first selected and processed in a processing block and then fed to a database that stores historical data as well as current data. The processing block may alternatively feed the data directly to a logic block that uses various parameters with their respective weights. The logic block calculates the credit-worthiness of the financial entity using a plurality of algorithms.
As per another embodiment, the customer background parameters comprise demographics of the decision makers or executives of the financial entity. Another embodiment describes the business model parameters comprising the market segment, collaborations and dependencies. Yet another embodiment has the asset specific parameters quantifying the need and intended utilization of asset or a plurality of assets. Yet another embodiment describes micro-financial performance ;

metrics such as short term financial performance and bank transactions and; macro-financial performance metrics such as long term performance, repayment track record and credit rating of the financial entity etc. Projected performance parameters comprise projected earning, customer pipeline and growth projections.
Yet another embodiment of the invention includes the data model using optimal weightages for various parameters and is further optimized using historical data analytics making it adaptive in nature. Historical data analytics can use various machine learning algorithms, heuristics, expert systems and /or statistics or a combination thereof. Another embodiment of the invention includes a relational fuzzy modeling system that builds a relational fuzzy model for the selected financial parameters. The system based on the data model is further optimized for specific clusters, wherein the clusters are based on business needs including but not limited to business segment, scale and volume.
In a another aspect, the invention provides a program product stored on a recordable medium for predicting credit worthiness of a financial entity based on an enhanced adaptive data model and historical data analytics.
BRIEF DESCRIPTION OF THE DRAWINGS
These and other features of this invention will be more readily understood from the following detailed description of the various aspects of the invention taken in conjunction with the accompanying drawings in which: Figure 1 depicts a computer system comprising a credit worthiness modeling system.

Figure 2 depicts a flow diagram of the modeling system of Figure 1. Figure 3 depicts a fuzzy reference set to be used in the logic block step S2 of Fig. 2 Figure 4 depicts a fuzzy logic process to be used in the logic block step S2 of Fig. 2 Figure 5 depicts a degree of membership for scalar values relating to Fig. 3.
DETAILED DESCRIPTION OF THE INVENTION
Referring now to the drawings, Figure 1 shows a credit worthiness modeling system 20 embodied as a program product in a computer system 10. As described in further detail below, credit worthiness modeling system 20 processes a set of financial data 26 and generates one or more credit worthiness models 28 of a financial entity. To accomplish this, credit worthiness modeling system 20 includes a data selection system 22 and a relational fuzzy modeling system 24.
In general, computer system 10 may comprise, e.g., a desktop, a laptop, a workstation, etc. Moreover, computer system 10 could be implemented as part of a client and/or a server. Computer system 10 generally includes a processing unit 14, memory 12, a bus, input/output (I/O) interfaces 16, external devices/resources and storage. The processing unit 14 may comprise a single processing unit, or be distributed across one or more processing units in one or more locations, e.g., on a client and server. Memory 12 may comprise any known type of data storage and/or transmission media, including magnetic media, optical media, random access memory (RAM), read-only memory (ROM), a data cache, a data object, etc. Moreover, memory

12 may reside at a single physical location, comprising one or more types of data storage, or be distributed across a plurality of physical systems in various forms.
I/O interfaces 16 may comprise any system for exchanging information to/from an external resource. External devices/resources (not shown) may comprise any known type of external device, including speakers, a CRT, LED screen, hand-held device, keyboard, mouse, voice recognition system, speech output system, printer, monitor/display, facsimile, pager, etc. A bus may be included to provide a communication link between each of the components in the computer system 10 and likewise may comprise any known type of transmission link, including electrical, optical, wireless, etc. Although not shown, additional components, such as cache memory, communication systems, system software, etc., may be incorporated into computer system 10.
A set of financial data 26 may be embodied in any type of storage system (e.g., a relational database, etc.) and may include one or more storage devices, such as RAM, ROM, a magnetic disk drive and/or an optical disk drive. Database can also be distributed across, for example, a local area network (LAN), wide area network (WAN) or a storage area network (SAN) (not
shown). Thus, The set of financial data 26 may be a database which could have some or all
of their data stored remotely over a distributed network, thereby allowing for the pooling of resources and information.
Such a network could be any type of network such as the Internet, a local area network (LAN), a wide area network (WAN), a virtual private network (VPN), etc. Communication could occur via a direct hardwired connection (e.g., serial port), or via an addressable connection that may

utilize any combination of wireline and/or wireless transmission methods. Moreover, conventional network connectivity, such as Token Ring, Ethernet, WiFi or other conventional communications standards could be used. Still yet, connectivity could be provided by conventional TCP/IP sockets-based protocol. In this instance, an Internet service provider could be used to establish interconnectivity. Further, as indicated above, communication could occur in a client-server or server-server environment.
It should be appreciated that the teachings of the present invention could be offered as a business method on a subscription or fee basis. For example, a computer system 10 comprising credit worthiness modeling system 20 could be created, maintained, supported and/or deployed by a service provider that offers the functions described herein for customers. That is, a service provider could offer the process of generating regulation models, e.g., as an application service provider.
Referring now to Figure 2, a flow diagram depicting the overall operation of the credit worthiness modeling system is shown. Various exemplary input data for the data model are depicted. Asset Specific Parameters 109 may include quantified evaluation of depreciating assets and may also include quantified evaluation of business need for at least one asset. Quantified socio-geographical parameters 103 may include indirect social or indirect professional connection. Other data may include interdependent multivariate analysis of various financial parameters. The data model further may include parameterized customer background 105 and business model parameters 107. The data model further includes financial performance metrics including but not limited to macro-financial performance indicators 101 and micro-

financial performance indicators 111. The macro-financial performance indicators 101 and micro-financial performance indicators 111.may both further include temporal financial performance indicators. The data model further may include projected performance parameters 113 for the financial entity.
In this first step SI of processing block, the relevant data is first selected from the data sources from 101 to 111 as described earlier and then processed in the processing block SI and then fed to a database S2 that stores historical data as well as current data. The processing block S2 may alternatively feed the data directly to a logic block S3 that uses various parameters with their respective weights. The logic block S3 then calculates the credit-worthiness of the financial entity using a plurality of algorithms. In a specific embodiment of the algorithm, at the step Logic block S3, a predefined model or a relational fuzzy model can be built using the representative subset to train the model. (Alternatively, a neural network based model could be built.).
If relational model is to be used, once the data is selected, relational fuzzy modeling system 24 fo Fig. 1, is used to build the model and test it on the same data. A predefined set of top results can then be reported. For the purposes of this disclosure, it is assumed that the reader has a basic understanding of the principals of fuzzy modeling, and more specifically relational fuzzy modeling.
Use of relational fuzzy modeling is different from rule-based modeling in that it is data driven as opposed to rule based modeling in which one needs to setup the rules based on domain knowledge. Thus, rule based fuzzy modeling has a variety of shortcomings in terms of updating

models, subjectivity inherent in the process, etc. Neural networks are better in that they are data driven and assume no knowledge of the system to be modeled. However, the output from a neural network is really a set of equations and is not easy for a scientist to interpret.
In fuzzy logic, a fuzzy set A on universe X is defined by the ordered pair where x
is the object on is called the membership function of A.The membership function can
be any value in the range of [0,1]. There are various types of fuzzy reference sets to describe the membership, one example is shown in Figure 3.
The foundation of fuzzy modeling and control is the concept of the fuzzy set, first described by Zadeh in 1965. In dealing with their everyday lives, people tend to reason about the world in qualitative terms. For example, they will talk about 'tall1 people, 'hot' water, Tast' cars, etc. In all these adjectives, there is understood to be considerable ambiguity, and a common reference frame between people is required for them to successfully communicate. For example, a 'fast' car in the early years of motoring might have been one that could reach 30 mph, but this is certainly not what we would understand by the term today. Similarly if we want to have machines that are able to process qualitative information in the form of rules, then we have to have some means of stating the terms of reference.
The natural way of defining a reference frame is to group things into sets; thus one would have a set of tall people, a set of hot water temperatures, a set of fast cars, and so on. However, conventional sets have a sharp cutoff between an element belonging to the set and it not belonging. This does not fit in well with what people actually mean by these qualitative;terms.

For example, if we specify a conventional set of tall people we might fix a set boundary at 1.8m. Now, with a conventional, or crisp, set someone who was 1.8m would be classed as tall, but someone whose height was 1.799m would not. This clearly is not the sort of meaning that a human would attach to the adjective 'tall'.
The fuzzy set is a way of dealing with real world ambiguity. A fuzzy set is a set with a fuzzy boundary, and the elements of the set belong to it with a variable grade of membership. This grade of membership is a number between zero and one, and it indicates how strongly a particular element belongs to the set (one indicating the strongest belonging and zero the weakest). For sets on continuous ranges (e.g., a range of real numbers) the membership of the set is defined by a membership function. For example, a fuzzy set of tall people might be defined so that everyone over 2.0 m in height belonged to the set with a grade of membership of one, everyone less than 1.5m with a grade of membership of zero, and people between 1.5m and 2.0m in height with some, varying, intermediate grade of membership.
Fuzzy membership functions can be of any shape that the designer decides is appropriate for the particular situation. Usually, however, one of a small group of functions is used to describe a fuzzy set, and the most commonly used nowadays is the triangular membership function. A triangular membership function only requires the position of three points to be specified: the leftmost edge of the set; the vertex of the set where the grade of membership is equal to one; and the rightmost edge of the set. The set can be 'opened' on one side simply by specifying a set boundary extending to infinity (or, anyway, a very large number!)

For modeling purposes, we need to define a group of sets that describe the range of each variable of interest. For example, to describe the heights of a group of people we might decide to use three fuzzy sets with the linguistic tags small, medium and large. The group of fuzzy sets which are specified for each variable are called the reference sets for that variable. The usual first step of processing in a fuzzy model or controller is the stage known as fuzzification. Here scalar values of the inputs are converted into possibility vectors. A possibility vector is simply a vector that describes the degree of membership of a particular input value in each of the reference sets defined for that input.
After fuzzification, the next stage is to process the fuzzy information in the possibility vectors through the rules describing the controller, or model, to form a fuzzy output possibility vector. To do this some compositional rules of inference are used to combine possibilities on both sides of any logical conditions. For example, a rule might say:
IF the person is tall AND the person is fit THEN the person can jump high. In this case the possibilities for the person being tall and for the person being fit have to be combined across a logical AND to give the overall degree of truth for the rule. There are a variety of different sets of compositional rules, but one of the most popular sets is: Across an AND: multiply the possibilities together; and Across an OR: add the possibilities together (but fix the maximum value at 1). Once the individual rules have been processed, we are left with a group of consequents from the rules giving values for the possibilities for the output lying in each of the output reference sets. Often several rules will havesthe same consequent (e.g., the person can jump high), but will have

fired at different strengths. In these cases it is necessary to combine the rules by taking the
maximum possibility for that consequent.
The final step is then to convert the possibility vector for the output into a scalar value that can
be used as, say, a signal to a control valve. Again, there have been several different methods
suggested for carrying out defuzzification, but the most popular is the fuzzy mean. An overview
of the fuzzy logic process is shown in Figure 4.
Relational fuzzy modeling of financial data has advantages over traditional fuzzy modeling and
can be implemented as follows. In analyzing financial data, the data is transformed from crisp
values to fuzzy values by fuzzification. A simplified set of financial data may appear as follows:
Let us assume that Percentage on-time payment record values are ranging between 0-100, and
for four time slots tl to t4, records for two financial entities are as follows:

tl t2 t3 t4
Entity 1 10 44 40 56
Entity 2 78 89 96 88
Fuzzification will replace each of the scalar values with a possibility vector that describes a degree of membership of each particular value. This is depicted in Figure 5, where a scalar value of 10 would be converted using three triangular reference sets L, M and H as a fuzzy vector {0.8, 0.2, 0.0}. This triplet means that a value of 10 is represented as being Low with a Grade of truth of "0.8", "0.2" as Medium and "0.0" as high.
Then, triplets of data are used to fill in the relational fuzzy matrix that is initially filled with zeros. Any fuzzy operational algorithm could be used for the fuzzified values to fill the

relational fuzzy matrix. This is done for all time stamps for the triplets. Then, the next triplet is used for updating this model as the base model.
Fuzzy relational systems differ from other fuzzy techniques in that a relational model has no explicit set of rules. Instead a relational array is used that maps every possible AND combination of the input reference sets to every output set. For example, consider a system with two inputs and a single output:
Output = f(Varl, Var2) If two reference sets, e.g., Low and High, are defined for each input variable, then there are four combinations of input reference sets, and each of these have to be mapped onto each of the similar two output reference sets. Thus the relational array will consist of eight elements such as:

Output set 1 Output set 2
Varl(seta) 0.2 0.5
Varl(setb) 0.8 0
Var2(set a) 0.2 0.1
Var4(set b) 0.7 0.5
The content of each element in the relational array is a number between zero and one, which
indicates how strong the particular relationship linked to that element is. A value of zero
indicates that the relationship does not apply, but values greater than zero indicate an increasing
strength in the relationship as they approach one. Since every possible relationship between the
inputs and outputs is included in the relational array, the values of the relational array elements
are what determine how the relational model behaves.

There are several ways of obtaining values for the relational array elements. One is simply to encode a rule-base into a relational format; in the simplest case this means that the array will hold values only of zero and one, since a rule either exists or it does not. Complications can arise with this approach, however, since rule antecedents are usually formulated with more flexibility than the relational model structure allows for.
Identification of the relational array directly from process input-output data is also possible, and this is the primary advantage that relational modeling has over its rule-based counterpart. There have been several identification algorithms proposed, but one of the best for noisy systems is that of Ridley, J.N., Shaw, I.S., and Kruger, J.J., "Probabilistic fuzzy model for dynamic systems," Electronic Letters, 24, (1988), pp 890-892, hereinafter "RSK." The RSK algorithm for this identification technique is:

R(sl,...,sn,s) = {SUM from {k=l} to N f_{sl,...,sn,k}. Y(s)k} over {SUM from {k=l} toN
f_{sl,...,sn,k}}
where,
R(sl ,..., sn , s) = An entry in the relational array
sn = The reference set index for the nth input
s = The reference set index for the output
N = Total number of samples of input-output data
i
fsl, ... ,sn, = The product of the input possibilities in reference sets

Yk = Possibility vector for the output (at sample k)
The key to the strength of this algorithm in dealing with noisy I/O data is the "f-factor," which gives a measure of the frequency of occurrence of each combination of inputs. The contribution of any particular sample of I/O data is thus weighted according to how frequently, and how strongly, that particular input combination is seen in the data. So, single examples of bad data are unlikely to have a significant effect on the values stored in the relational array.
It is understood that the systems, functions, mechanisms, methods, engines and modules
described herein can be implemented in hardware, software, or a combination of hardware and
software. They may be implemented by any type of computer system or other apparatus adapted
for carrying out the methods described herein. A typical combination of hardware and software
could be a general-purpose computer system with a computer program that, when loaded and
executed, controls the computer system such that it carries out the methods described herein.
Alternatively, a specific use computer, containing specialized hardware for carrying out one or
more of the functional tasks of the invention could be utilized. In a further embodiment, part of
all of the invention could be implemented in a distributed manner, e.g., over a network such as
the Internet.
The present invention can also be embedded in a computer program product, which comprises all
the features enabling the implementation of the methods and functions described herein, and
which - when loaded in a computer system - is able to carry out these methods and functions.
Terms such as computer program, software program, program, program product, software, etc.,

in the present context mean any expression, in any language, code or notation, of a set of instructions intended to cause a system having an information processing capability to perform a particular function either directly or after either or both of the following: (a) conversion to another language, code or notation; and/or (b) reproduction in a different material form. The foregoing description of the invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed, and obviously, many modifications and variations are possible. Such modifications and variations that may be apparent to a person skilled in the art are intended to be included within the scope of this invention as defined by the accompanying claims.

CLAIMS We Claim:
1. A system for predicting credit worthiness of a financial entity using at least one
parameter selected from a list comprising quantification of:
- evaluation of depreciating assets;
- indirect social or indirect professional connection;
- interdependent multivariate analysis of various financial parameters; and
- evaluation of business need for at least one asset.
2. The system of claim 1, further comprising a combination of parameters selected from a
set including parameterized customer background, business model parameter, asset
specific parameter, quantified socio-geographical parameter, macro-financial
performance indicator, micro-financial performance indicator, temporal financial
performance indicator and projected performance parameters for the financial entity.
3. The system of claim 2, wherein
the customer background parameter comprising demographics of the decision makers or executives of the financial entity;
the business model parameter comprising the market segment, collaborations and dependencies;

the asset specific parameter quantifying the intended utilization of asset or a plurality of assets;
the micro-financial performance metric comprising short term financial performance and bank transactions;
the macro-financial performance metric comprising long term performance, repayment Track Record and credit rating of the financial entity;
the temporal financial performance indicator comprising long term bank transactions; and
the projected performance parameters comprising projected earning, customer pipeline and growth projections.
4. The system of claim 3, further comprising:
A processing block for selecting relevant data, processing the selected data and
4
feeding the selected data to at least one of a database and a logic block;
The database for storing at least one of historical data and current data; The logic block configured to:
-receive weights for various parameters;
-receive data from at least one of the database and the processing block; and
-calculate the credit-worthiness of the financial entity using a plurality
of algorithms.

5. The system of claim 4, wherein the database is at least one of relational database and hierarchical database and further the plurality of algorithms comprises machine learning algorithms, heuristics, expert systems and statistics.
6. A system for predicting credit worthiness of a financial entity that
uses at least one parameter selected from a list comprising quantification of:
- evaluation of depreciating assets;
- indirect social or indirect professional connection;
- interdependent multivariate analysis of various financial parameters; and
- evaluation of business need for at least one asset; and
uses a relational fuzzy modeling system that builds a relational fuzzy model using financial data of the financial entity.
7. The system of claim 6, further comprising a combination of parameters selected from a
set including parameterized customer background, business model parameter, asset
specific parameter, quantified socio-geographical . parameter, macro-financial
performance indicator, micro-financial performance indicator, temporal financial
performance indicator and projected performance parameters for the financial entity.

8. The system of claim 7, further comprising:
A processing block for selecting relevant data, processing the selected data and feeding the selected data to at least one of a database and a logic block;
The database for storing at least one of historical data and current data; The logic block configured to:
-receive weights for various parameters;
-receive data from at least one of the database and the processing block; and
-calculate the credit-worthiness of the financial entity using the relational fuzzy model.
9. The system of claim 8, wherein the relational fuzzy model includes a relational array built using an identification algorithm defined as: R(sl,...,sn,s) = {SUM from {k=l} to N f_{sl,...,sn,k}. Y(s)k} over {SUM from {k=l} to N f_{sl,...,sn,k}} where,
R(sl ,..., sn , s) = An entry in the relational array
sn = The reference set index for the nth input
s = The reference set index for the output
N , = Total number of samples of input-output data
fsl, ... ,sn, = The product of the input possibilities in reference sets

Yk = Possibility vector for the output (at sample k).
Possibility vector for the output (at sample k).
10. A program product stored on a recordable medium for predicting credit worthiness of a financial entity, comprising:
A processing block for selecting relevant data, processing the selected data and feeding the selected data to at least one of a database and a logic block;
The database for storing at least one of historical data and current data; The logic block configured to:
-receive weights for various parameters;
-receive data from at least one of the database and the processing block; and
-calculate the credit-worthiness of the financial entity using the relational fuzzy model. wherein the selected data comprises at least one selected from a list comprising quantification of: evaluation of depreciating assets, indirect social or indirect professional connection, interdependent multivariate analysis of various financial parameters, evaluation of business need for at least one asset, parameterized customer background, business model parameter, asset specific parameter, quantified socio-geographical parameter, macro-financial performance indicator, micro-financial performance indicator, temporal financial performance indicator and projected performance parameters for the financial entity; and

wherein the relational fuzzy model includes a relational array built using an identification algorithm defined as:
R(sl,...,sn,s) = {SUM from {k=l} to N f_{sl,...,sn,k}. Y(s)k} over {SUM from {k=l} to N f_{sl,...,sn,k}} where,
R(sl ,... sn , s) = An entry in the relational array
sn = The reference set index for the nth input
s = The reference set index for the output
N = Total number of samples of input-output data
fsl,... ,sn, = The product of the input possibilities in reference sets
Yk = Possibility vector for the output (at sample k).
Possibility vector for the output (at sample k).