prasath's img

Prasath Murugesan

I'm a Researcher and Data Science student.
I'm looking for Data Science roles from December 2022.

About Me

My name is Prasath M. I’m currently pursuing final year of 5-year Integrated M.Sc Data Science offered by the Department of Applied Mathematics and Computational Sciences at PSG College of Technology, Coimbatore. Recently I founded and lead the Computational Data Science Club at PSG College of Technology. I’m currently exploring knowledge graphs and am inclined to build context-driven AI systems which has wide range of applications in the industry.

Languages

Python, R

100%

MySQL, PostgreSQL

100%

C, C++

100%

Java

100%

Libraries proficient

Numpy, Pandas

100%

Scikit, Tensorflow

100%

NLTK, Spacy

100%

NetworkX, Stellargraph

100%

Matplotlib, Seaborn

100%

BeautifulSoup, Streamlit(UI)

100%

Areas of Interest

Machine Learning

100%

Graph Models & knowledge Graphs

100%

NLP

100%

Data Analytics & Mining

100%

Graph Mining

100%

Explainable AI

100%

Education

2018-2023

PSG College of Technology, Coimbatore, India.

Master's in Data Science (M.Sc Integrated).
CGPA till VIII Semester - 8.69 CGPA.

2012-2018

Jawahar Higher Secondary School (CBSE), Neyveli, India.

Computer Science & Mathematics.
Graduated Grade XII with 93.2%.
Graduated Grade X with 10 CGPA.

Research Experience & Internship

August 2020 - September 2022

Collective Classification Models for Data Without Explicit Linkage.

Prasath Murugesan, Shamshu Dharwez

In this paper, we consider the task of text categorization as a graph classification problem without explicit linkage. We demonstrate that even on low volumes of labelled data, the graphical inference models built using the proposed methodology outperform the traditional models and also a few other models of the collective classification family.

May 2021 - November 2021

Research & Development Intern, AI Institute of University of South Carolina.

During the internship, I was managing the team and developed on the problem of “Generalization in TextWorld Commonsense” where we leveraged ConceptNet Knowledge graph and KagNet to improve generalization and explainability in RL agents using the TextWorld Commonsense Sandbox environment.

My Projects

News Recommendation System at Scale

A news recommendation system built using MIcrosoft News Dataset (MIND) containing 750K users and 17.5M click events. It uses Collaborative Filtering, KNN and Graph Projections which helps to recommend similar news articles based on user preferences and enable rank-ordered recommendation queries personalized to each user.

Music Meta Brain

A contextual information retrieval system which leverages knowledge graph built in an unsupervised setting which helps to extract relevant information for music search queries. It is helpful in extracting relevant and similar documents to the queries generated.

Sentiment Detection Web app

It is an end-to-end interactive web app used to predict and visualize probability distribution of emotions using Multinomial Logistic Regression Pipeline.

Balancing the imbalanced

Resolved the imbalance in credit card fraud detection dataset which is highly imbalanced containing 0.01% of fraud transactions. Built a Random Forest model to construct a comparison flow chart for 8 resampling techniques. It helps to reduce the number of False Positives and leads to better precision and recall measures.

Time Series Analysis on beverage drink consumption

Analyzed the trend in beer consumption of Australians during 2000 - 2018 using R. Trailed to find patterns and modeled exponential smoothing and trend methods. The beer sales were increasing annually with a seasonal decrease in mid-quarters every year.