Arinjoy Basak

A Graph Based Feature Selection Algorithm Utilizing Attribute Intercorrelation

Arinjoy Basak, Asit Kr. Das

The 7th IEEE Annual Information Technology, Electronics and Mobile Communication Conference (IEEE - IEMCON 2016), University of British Columbia, Vancouver, Canada, October 2016

Recently, every enterprise generates large volumes of on a regular basis. Complex data mining and analysis techniques are used to feasibly analyse high dimensional data. Feature selection aids in this by providing a re- duced representation of this data while maintaining integrity. We propose a graph-based feature selection algorithm utilizing feature intercorrelation to construct a weighted attribute graph, from which attributes are iteratively removed to construct the reduct based on a scoring scheme. Disconnectivity of the graph serves as the point of termination for our algorithm. The performance of our algorithm on real valued and discretized datasets is evaluated statistically by generating the Receiver Operator Characteristic (ROC) curve for each reduced dataset, and by measuring accuracies for classification training tasks, for the datasets reduced by our method.

Check out the paper here

[Report Excerpt] Data Analytics for IITBombayX (Based on OpenEdx InSights) - Detection of difficulty regions in lecture videos of students

Arinjoy Basak

Ekalavya 2015, Department of Computer Science, IIT Bombay

This part of the project report talks about the project work done regarding the development of models for the detection of difficulty regions in videos based on the students behaviour, recorded through the log events in IITBombayX. Such inferences can be determined for a video, in a particular course, based on the students' activities on the lecture video – and inform the same to the course instructors or coordinators, who would then take appropriate steps to address such a problem. We provide a basic outline of our idea to deal with the problem and subsequently, we describe how we had designed the data model and the functionality both at the processing and visualization level for a final implementation of the analytics module in the scenario of Big Data. After the final implementation and testing, the module was integrated with the OPENedX Insight analytics system for IITBombayX, and made available for use by the course instructors participating in the Blended MOOCs model. The entire work of this part was completed in 3 weeks of the internship.

Check out the excerpt of my work here

Check out the FULL report here

Achieving Data Survivability and Confidentiality in Unattended Wireless Sensor Networks

Arpan Sen, Shrestha Ghosh, Arinjoy Basak, Harsh Parshuram Puria, Sushmita Ruj

The 29th IEEE International Conference on Advanced Information Networking and Applications (AINA-2015) , Gwangju, Korea, March 2015

In Unattended Wireless Sensor Networks (UWSNs) the nodes are subjected to hostile environment for sensing critical data. Due to the unattended nature of the network the sink is not always present. Hence, the nodes in the network are required to function in a distributed way in order to ensure Data Survivability and Data Confidentiality. In this work we address these two issues. We have proposed algorithm (s) to ensure Data Survivability by encryption and data replication. We propose a simple scheme for key management which ensures confidentiality by sharing the key among various nodes in the network so that the adversary cannot read the data by compromising a node in the network. We have compared our scheme with the existing ones, both mathematically and by simulations. Analysis shows that our scheme performs better in terms of overheads and efficiency.

Check out the paper here