Developed a model to detect Parkinson's Disease using Multidimensional Voice Program (MDVP) Analysis parameters. Analyzed 195 MDVP data records, differentiating between 31 individuals with Parkinson's and 8 without. Explored 33 quantitative voice parameters, utilizing statistical hypothesis testing to establish predictive capabilities. Addressed challenges like class imbalance and feature collinearity through data sampling and careful parameter selection. Employed exploratory data analysis, identifying key parameters for modeling. Leveraged machine learning techniques, including Gradient Boosted Classification, to achieve a 93.9% recall score, indicating high accuracy in identifying Parkinson's cases.

Click here to learn more on my github

Conducted topic exploration across 33 car brand subreddits to discern underlying discussions. Utilized Natural Language Processing techniques, including Non-Negative Matrix Factorization (NMF) and Latent Dirichlet Allocation (LDA), to extract prevalent topics from ~200,000 comments. Processed data by removing stopwords and lemmatizing words for clarity. Explored topics like brand-specific discussions, pic threads, and unique conversations in specific subreddits. Achieved insights into varying patterns of discussions across different automotive communities. Results visualized through topic models provided valuable understanding of community engagement.

Click here to learn more on my github