Claims:1) Dynamic, low cost, mobile telecommunication tower placement system comprising:
An Android Application that acts as a data aggregation unit (DAU), which collects telecommunication data from the smartphones.
A centralized server that collects data from several DAUs and
A tower placement scheme that utilizes several algorithms of data science on the collected data.
said Android Application collects the following parameters for data analysis:
Geolocation data (latitude and longitude in radian)
Signal Strength (in dBm)
Carrier name
Network Type
Date and Time
said Android Application also
Records data at frequent intervals to keep a healthy sample size, but doing so with nominal battery consumption.
Appends collected data in a single file that is transmitted to the centralized server at fixed intervals, on availability of internet connection.
Centralised server also
Ignores anomalous readings, or readings with some parameters not recorded.
Ignores redundant entries with same time stamp and same geolocation value.
Checks if any reading has network type “Wifi” and substitutes it with the last network type detected that is not “Wifi”.

2) A tower placement scheme as claimed in claim 1, which comprises three steps, namely, clustering, classification and positioning.

3) Clustering as claimed in claim 2, involves Mean Shift Clustering on the geolocation data collected by the DAUs over a period of time from several users. Mean Shift Clustering provides Nc number of clusters, and a centroid for each cluster.

4)Classification, as claimed in claim 2, involves classifying clusters obtained from Clustering, as claimed in claim 3, into two zones each, a zone with good signal strength and another zone with bad signal strength.

5) Classification, as claimed in claim 2, is additionally provided the following inputs:
Sg ,which is the minimum Signal Strength above which the signal strength is good for normal usage by a user.
Sb ,which is the maximum Signal Strength below which the signal strength is bad for normal usage by a user.
Nmin, ,which is the minimum number of users per telecommunication tower, that the carrier choses to have, so that the installation is economically feasible.
Rsignal_min,which is the Minimum Desirable Cluster Signal Ratio, defined in claim 4.

6)Classification , as claimed in claim 2, involves calculating the following parameters for every cluster:
Ng , which is the number of recorded locations where signal strength recorded is greater than Sg
Nb , which is the number of recorded locations where signal strength recorded is lesser than Sb.
a parameter known as a cluster signal ratio Rsignal_i which is the ratio of Ng to Nb for the ith cluster.

7) Classification, as claimed in claim 2, is performed only on those clusters where Rsignal_i Sg to the number of points (Nb) in the cluster where S(i,j) < Sb.
Rsignal_i = Ng / Nb
From the Nc clusters, for further steps, we consider only those clusters where Ni > Nmin and Rsignal_i < Rsignal_min . These clusters are added to a set S_svm and each cluster in this set is further divided into two types of zones, good signal strength zones and bad signal strength zones. This is done using Support Vector Machine (SVM) Classification. SVM classifies a given set of data into two discrete sets and using SVM helps in concentrating the focus completely on the region which actually requires a new tower placement and installation.

3. Positioning: We now concentrate on the weak signal zone obtained from classification. We then perform K-means algorithm to find the centroid (cluster center) which is the possible candidate site for tower placement.

For a given cluster, the K-Means Algorithm works in the following way:

a. Randomly select ‘k’ number of cluster centers.
b. Calculate the distance between each data point and each cluster center.
c. To every data point, assign that cluster center which is closest to that point.
d. Calculate the new cluster center.
e. Again, calculate the distance between each data point and the newly obtained cluster center.
f. If none of the data points is reassigned to a new cluster center, stop; otherwise repeat steps (c) through (f).

We then consider the given candidate site so as to see the geographical suitability (Water Bodies, hilly areas etc. are neglected and land availability.