Projects
NLP-Based Classification and Severity Assessment of Anxiety and Mental Health Conditions
This project explores the application of Natural Language Processing (NLP) techniques for symptom classification and severity assessment in mental health contexts. The study consists of two main tasks: binary classification to identify the presence of anxiety-related symptoms, and multiclass classification to categorize symptom severity. The methodology includes comprehensive text preprocessing, exploratory data analysis using word clouds and topic modeling, and the implementation of various machine learning algorithms. For the binary classification task, Random Forest with Grid Search CV emerged as the top-performing model, achieving 93% accuracy. In the multiclass severity classification task, Naive Bayes with feature engineering proved most effective, attaining 91% accuracy. The project demonstrates the potential of NLP and machine learning in automating the analysis of mental health-related text data, which could have significant implications for early symptom detection and severity assessment in clinical settings. The findings suggest that these techniques can provide valuable insights for mental health professionals, potentially improving diagnosis and treatment strategies.
Sentiment Classification using Advanced Data Analytics for ChatGPT
Explored advanced data visualization techniques to analyze ChatGPT discussions on Twitter. Investigated public perceptions of ChatGPT using Python and Weka and conducting sentiment analysis.
University Grading System Enhancement
As a Research Assistant, contributed to improving the university grading system by introducing a novel relative grading approach. Applied data-driven methodologies to enhance precision and fairness.
Implementing Homomorphic Encryption on Datasets
In this project, I explored the use of homomorphic encryption to securely perform computations on encrypted data without decryption. I utilized PYFHEL and Concrete-Numpy libraries to encrypt small datasets and execute numpy operations on the encrypted data. The study focused on evaluating the feasibility and performance of applying machine learning algorithms on homomorphically encrypted data compared to unencrypted datasets. Despite the added complexity and increased processing requirements, PYFHEL facilitated the implementation of fully homomorphic encryption, demonstrating its potential for secure data processing.