Soumyajyoti Banerjee

This project is a location dimension services based on OSRM (open street maps https://www.openstreetmap.org/) using historical hyperlocal data of eatfit pilot service. This service incorporate current travel data using active learning. Dividing the area based on lat-long square and incorporating the sample based current data makes this framework 92% accurate with respect to Google Maps. This reduce the cost of distance service to 10% of the Google API cost. As it require the AWS box only… Show more

This project is a location dimension services based on OSRM (open street maps https://www.openstreetmap.org/) using historical hyperlocal data of eatfit pilot service. This service incorporate current travel data using active learning. Dividing the area based on lat-long square and incorporating the sample based current data makes this framework 92% accurate with respect to Google Maps. This reduce the cost of distance service to 10% of the Google API cost. As it require the AWS box only.
This parallel service helps to boost hyperlocal service of cure fit with better accuracy. Show less

Business Problem:
Objective of this exercise was to parse standardize US address and detect/tag parts of the US address like pre direction, post-direction, house number, street name, street type, suite name suite number, OCC number and type.
Abstract:
Main objective of this programme was to analyse and find an intelligent way to parse US address as well tag the components in a faster way. Which will help to enrich address information as well help in better informative address… Show more

Business Problem:
Objective of this exercise was to parse standardize US address and detect/tag parts of the US address like pre direction, post-direction, house number, street name, street type, suite name suite number, OCC number and type.
Abstract:
Main objective of this programme was to analyse and find an intelligent way to parse US address as well tag the components in a faster way. Which will help to enrich address information as well help in better informative address matching between two entity . After a rigorous analysis of US address we came up with a really fast and intelligent technique to tag the US address using a back population based algorithm using it’s NLP based features.
Accuracy : 95.4%
Framework: Spark ,
Language: Python
Business Impact: It has enabled the huge capacity of address based Merchant Identification and gave an opportunity to do high speed solr indexing based on address based component . Which became a integrated part of Location based repository of US based merchant. That has huge impact on merchant hierarchy based developments.
Show less

Objective of this project was to clean the Raw merchant names come through the transactions in run time.
Problems: 1. Identify the junk in the raw name to clean the raw name.
2. Identify the intelligent way to build the tree with huge amount of data to use that to clean raw names.
3. Use the tree as functionalities to clean the String.

We achieved a sustainable and intelligent solution that can eliminate the contextual junk from the string using… Show more

Objective of this project was to clean the Raw merchant names come through the transactions in run time.
Problems: 1. Identify the junk in the raw name to clean the raw name.
2. Identify the intelligent way to build the tree with huge amount of data to use that to clean raw names.
3. Use the tree as functionalities to clean the String.

We achieved a sustainable and intelligent solution that can eliminate the contextual junk from the string using the tree of merchant names.

Technologies: Hadoop, Spark, Apache Parquet, Hive
Language: Python Show less

We built designed and built the framework to simplify the 3-Party Model of Doctor, Patient and Pharmacist in Health domain making their life simpler to deal with day to day health issues. We developed 3 applications: One in Android Phone, Second one a Java Spring Framework and Last one a Python Framework using technologies in the domain of Image Processing, Natural Language Processing, Android, Real Time Payment, Real Time Social Network Feed. Our framework can bring down the operating time of… Show more

We built designed and built the framework to simplify the 3-Party Model of Doctor, Patient and Pharmacist in Health domain making their life simpler to deal with day to day health issues. We developed 3 applications: One in Android Phone, Second one a Java Spring Framework and Last one a Python Framework using technologies in the domain of Image Processing, Natural Language Processing, Android, Real Time Payment, Real Time Social Network Feed. Our framework can bring down the operating time of all parties by more than significant amount.


Used Technologies:
#Android Development Using Java
#OCR framework integrated with database
#Front end for physicians using Java Swing
#Push payments using Visa API
#Automation payment, mailing bill/prescription, and warehouse management for medical shops using Java Show less

This is a part of huge paradigm shift from a Abnitio based framework to hadoop framework, In this project main constrain was to deal with huge amount of data and to match the high expectation of the speed.

Technology: Hadoop, Spark
Language: Java, Python
We matched the expectation using intelligent use of that huge amount of data and optimizing lookup time.

Inspiration
------------------
Long waits in toll booth are the most irritating thing that one can face while driving. Long queues and rummaging for change may easily sour one's sweet driving experience.
------------------
What it does
------------------
Our app focuses on eliminating these challenges via push payments at the predefined toll booth making them self serviceable resulting in a hassle free automated solution to the toll problem. The authorities get a transparent… Show more

Inspiration
------------------
Long waits in toll booth are the most irritating thing that one can face while driving. Long queues and rummaging for change may easily sour one's sweet driving experience.
------------------
What it does
------------------
Our app focuses on eliminating these challenges via push payments at the predefined toll booth making them self serviceable resulting in a hassle free automated solution to the toll problem. The authorities get a transparent system to manage fund collections.
------------------
How we built it
------------------
We used card to card push payment using Two Way SSL Authentication for a hassle free payment. We used Google Cloud Messaging server to send push messages of successful transaction to toll plaza and upstream message from toll plaza to send the QR Code to the client. We crawled web using python to collect toll data. Platform is Android
------------------
Challenges we ran into
------------------
Sending Upstream messages;Time;Integration of Various Modules
------------------
Accomplishments that we're proud of
------------------
We collected proper and full data from Web using a self built automated script. We remove the use of cash from the whole ecosystem to make the payment transparent to the financial authorities.
------------------
What we learned
------------------
Lots of Android Classes and APIs
------------------
What's next for ParkIt
------------------
The generality of the solution can make it viable for parking problem as well and the destination data can be used to make recommendation to the user about restaurants and events using social media analytic and other publicly available APIs'.
------------------
Built With
Android;Google Cloud Messaging;Python;VISA Push Payment;Zomato APIless Show less

Business Problem:
Objective of this exercise was to connect merchant information of transactional data with the in house existing merchant information.
Abstract:
Main objective of this project was to find an intelligent way of connecting transactional merchant information with existing sources of merchant information. Main challenge was to match distorted Merchant name coming in the transactional data with the existing sources. Based on name and geographical data We developed a smart… Show more

Business Problem:
Objective of this exercise was to connect merchant information of transactional data with the in house existing merchant information.
Abstract:
Main objective of this project was to find an intelligent way of connecting transactional merchant information with existing sources of merchant information. Main challenge was to match distorted Merchant name coming in the transactional data with the existing sources. Based on name and geographical data We developed a smart NLP based feature powered machine learning based technology to achieve the target.

Name Matching (Based on Distorted String) :
----------------------------------------------------------------
We have used NLP based features and Machine Learning based technology to predict the most matched name from a huge data set of merchant names using intelligent techniques.

Address Matching (Based on Approx. Address)
In address matching we used USPS codes and other location based services to standardize the address and matched based on location based features with several intelligent techniques.

Based on this two features we calculated a cumulative confidence score of matching and returned the merchant with complete information.
Accuracy : 81.3%


Technology: Hadoop, Spark, Hive, Apache Parquet, Spark Sql, Machine Learning tools,
Language: Python

Business Impact: It has enabled smart merchant identification in several downstream procedures. Not only helped to check issuer given data but also improved several real time procedures with higher accuracy. Show less

Ever increasing popularity of social networks in our daily life leveraged us to solve and
support several severe problems. Here in this paper we tried to recommend a user centric community based on users‘ interest topics, location and relationship network. This can solve several significant socio-psychological issues and also will help us to understand
information propagation. Several research works have been done on the user recommendation
and community detection but novelty of our… Show more

Ever increasing popularity of social networks in our daily life leveraged us to solve and
support several severe problems. Here in this paper we tried to recommend a user centric community based on users‘ interest topics, location and relationship network. This can solve several significant socio-psychological issues and also will help us to understand
information propagation. Several research works have been done on the user recommendation
and community detection but novelty of our approach is that we are considering user’s network structure to calculate the proximity between any two users. In this recommendation procedure only the users who are interested in socialization are recommended. This has been predicted based on several features of a user profile. In this recommendation procedure we tried
to recommend human users only by eliminating bot users and cyborgs from the recommendation set. We tried to use several properties to analyse a user. We have shown that these properties can be leveraged to improve the performance and effectiveness of the user eccentric community recommendation in online social networks. We have conducted several experiments
using Twitter data to observe the perfection and performance. Show less

This project was a part of a competition arranged by Microsoft Code.fun.do. Here we developed an Windows phone 8 app that aims towards providing better security to women.The app uses the cloud and gps based services to perform this task. It was greatly appreciated by judges

This Music Player Play & Download two work done in the same time.

This Project deals with the demand prediction of Eat.fit using the data of (t-2). This project was mostly divided into two parts.
1. Singles Order Prediction
2. Pack/Subscription Level Prediction

For Singles Prediction developed one relative probability based time series model using ARIMA and computed demand for each item available on that day.
For Packs data developed a regression model based several derived features like Pause Count, Re Active Count, Cancel Count, Sell Amount… Show more

This Project deals with the demand prediction of Eat.fit using the data of (t-2). This project was mostly divided into two parts.
1. Singles Order Prediction
2. Pack/Subscription Level Prediction

For Singles Prediction developed one relative probability based time series model using ARIMA and computed demand for each item available on that day.
For Packs data developed a regression model based several derived features like Pause Count, Re Active Count, Cancel Count, Sell Amount using time series based model (ARIMA, ARIMAX).

#This Model Reduced The Wastage on an average from 40-45% to 11-15%. Show less

This project deals with extracting information from visa net transaction. This service enable the function of extracting street address, phone number, store number and proper merchant name. This process has been built based on most frequent tokens and their importance and regex. This Solution has been built on Map Reduce framework .

This project deals with merging Merchant data from different sources. This solution built on Map Reduce framework powered by solr engine for faster data retrieval. This project involves multi level name matching, optimized geo matching and street address matching to achieve the desired goal. Developed a smart key to cluster merchants and reduce search space to find appropriate match. This is a huge effort to make the in house data sources more consolidated as well as consistent for more… Show more

This project deals with merging Merchant data from different sources. This solution built on Map Reduce framework powered by solr engine for faster data retrieval. This project involves multi level name matching, optimized geo matching and street address matching to achieve the desired goal. Developed a smart key to cluster merchants and reduce search space to find appropriate match. This is a huge effort to make the in house data sources more consolidated as well as consistent for more information extraction and different analysis and report publication purpose. Show less

Part of http://www.uni-weimar.de/medien/webis/events/pan-15/pan15-web/author-profiling.html
Authorship analysis deals with the classification of texts into classes based on the stylistic choices of their authors. Beyond the author identification and author verification tasks where the style of individual authors is examined, author profiling distinguishes between classes of authors studying their sociolect aspect, that is, how language is shared by people. This helps in identifying profiling… Show more

Part of http://www.uni-weimar.de/medien/webis/events/pan-15/pan15-web/author-profiling.html
Authorship analysis deals with the classification of texts into classes based on the stylistic choices of their authors. Beyond the author identification and author verification tasks where the style of individual authors is examined, author profiling distinguishes between classes of authors studying their sociolect aspect, that is, how language is shared by people. This helps in identifying profiling aspects such as gender, age. Our main task was to detect the gender and age of the author using several NLP techniques and Machine learning techniques. Show less

This project is concerned with making a mobile cloud using wifi-direct, to use it for high resource consuming applications like augmented reality. This project enabled to communicate among devices and divide resources intelligently for high resource consuming task.

Part of our course we did this project,. This was related to making an inconsistent library data set to a consistent data set by using several noise reduction and outlier detection methods. And we achieved a decent level of clarity in that data set. We used some Natural Language processing and N-gram based approach to make it more consistent.