Roshan Kumar

• Implemented a matrix multiplication algorithm (size 10000 x 10000) using message passing
interface. Achieved in reducing the computation time by matrix size to number of processor ratio
approach.
• The program was tested using 27 processors available in the CSC lab in IIT Delhi .

• Built a prototype for Search Engine based on cosine similarity and prestige ranking of web pages.
• Implemented a web crawler, HTML parser and indexer to index 1000 web pages.

• Implemented an Artificial Neural Network (ANN) System in MATLAB for time series analysis of
temperature variation of various months. Implemented PSO-RBF and MNN for training the ANN

• Created software to manage XML files for multiple users with read, write and access restriction on
files using JSP as the user interface. Innovated an XQuery Interpreter to perform queries and added
an Activity log to keep track of all users.

Objective of the project was two fold:
1)Predict whether or not a customer would donate
2)Predict the amount a person would donate

Built logistic regression model to predict whether a customer is likely to donate
Built multiple regression model to predict the donation amount of a person likely to make a donation
Identified top 1000 donors to target based on predicted donation amount
Built other models ridge and lasso regression and used cross validation to find the best… Show more

Objective of the project was two fold:
1)Predict whether or not a customer would donate
2)Predict the amount a person would donate

Built logistic regression model to predict whether a customer is likely to donate
Built multiple regression model to predict the donation amount of a person likely to make a donation
Identified top 1000 donors to target based on predicted donation amount
Built other models ridge and lasso regression and used cross validation to find the best model Show less

The project objective was to find the correct chain of emails between questions and answers, create an FAQ and propose template answers for the same.

Analyzed more than 8000 emails using the Text Mining package in R.
Created text corpora, removed stop words, performed stemming and clustering.
Created a matching algorithm based on name and subject of email. Also performed the temporal analysis on time stamp to find the sequence of mails.
Techniques used : tf-idf weights, term… Show more

The project objective was to find the correct chain of emails between questions and answers, create an FAQ and propose template answers for the same.

Analyzed more than 8000 emails using the Text Mining package in R.
Created text corpora, removed stop words, performed stemming and clustering.
Created a matching algorithm based on name and subject of email. Also performed the temporal analysis on time stamp to find the sequence of mails.
Techniques used : tf-idf weights, term frequency - inverse document frequency and word semantics (cosine similarity). Show less

• Implemented a "Recommender System" based on Collaborative and Knowledge based technique
• Improved the recommendation by combining the two techniques, making a Hybrid Recommender

• Implemented the Eigenface algorithm by performing principal component analysis (PCA) on a large set of images depicting different human faces.

• Implemented the comparison of finite element and finite difference solutions of one and two dimensional Burger’s Equation using Matlab. It has an extensive usage of Partial differential equations and finite element and finite difference schemes