Abhinav Arora

Worked on using collaborative filtering to develop a movie recommendation system for users on the given Netflix Prize Dataset. Implemented memory-based method & model-based methods to predict the ratings of movies for new users. Used Pearson's Correlation Coefficient (PCC) to account for user and movie bias in the ratings.In addition, performed bipartite clustering to compute user and movie clusters.

Implemented a general purpose dynamic memory allocator in C - malloc( ), free( ), calloc( ) and realloc( ). Built using segregated lists with a first fit algorithm and constant coalescing time.

Developed a text-mining based machine learning system to predict the user ratings of a business using the review text. Trained Logistic Regression and SVM models using various language based feature engineering techniques. The model that gave the best accuracy was chosen through cross-validation.

Built a search engine in Java to retrieve and rank documents using boolean, BM25 and Indri retrieval models. Implemented LETOR re-ranking(SVM-rank) and query expansion by pseudo-relevance feedback for better results. Conducted experiments to evaluate search quality using metrics like Mean Average Precision and Precision@N.
Technologies Used: Java, Lucene

Implemented a reinforcement Bipartite Clustering algorithm on news articles from the TDT4 Dataset. The algorithm produces both document clusters and word clusters simultaneously. Algorithm was evaluated using both external metrics (F-1 Score) and internal metrics (sum of cosine similarities of documents to the corresponding centroid).

The following 3 objectives were accomplished:-
1. Generation of Test Cases from UML 2.0 Sequence Diagrams.
2. Prioritization of Test Cases using Genetic Algorithm.
3. A GUI based tool was designed in C++ which generated the Test Suite from a Sequence Diagram

Lead Developer of the PHP based Web Placement Portal (with 3 other team members) to facilitate the strenuous task of Training and Placements in NSIT. It is being used by the Training and Placement Department of NSIT (http://tnp.nsitonline.in) since August 2010.

The basic objective was the elimination of paperwork for the placement procedure. It uses PHP, HTML, CSS, JavaScript, JQuery, Ajax & MySQL technologies. Implemented various features in the above technologies to provide a User… Show more

Lead Developer of the PHP based Web Placement Portal (with 3 other team members) to facilitate the strenuous task of Training and Placements in NSIT. It is being used by the Training and Placement Department of NSIT (http://tnp.nsitonline.in) since August 2010.

The basic objective was the elimination of paperwork for the placement procedure. It uses PHP, HTML, CSS, JavaScript, JQuery, Ajax & MySQL technologies. Implemented various features in the above technologies to provide a User Friendly system. Show less

Designed a classifier to predict the action of an individual based upon the activation of the brain voxels. The final classifier was a Voting classifier that was made from an ensemble of three classifiers. The three classifiers that we used were SVM with RBF Kernel, Logistic Regression classifier and Gradient-Boosted Trees. Also developed a regression model to impute missing voxel data using regression.
Technologies Used: Python and Scikit-Learn