📊 Analytics, Data Science, and AI
Projects in Machine Learning & Applied Data Science
Durham University & MIT Applied Data Science Program
This repository showcases hands-on data science projects completed during my studies at Durham University and the MIT Applied Data Science Program. Each project demonstrates key data science skills—data cleaning, modeling, visualization, and interpretation—across multiple domains such as finance, genetics, affective computing, and natural language processing.
⚠️ Some original datasets used are not publicly available due to licensing restrictions.
🔹 Affective AI: Facial Emotion Recognition
Facial Emotion Recognition Project
A deep learning model built with TensorFlow and Keras to classify facial emotions (happy, sad, neutral, surprise) using a dataset of over 20,000 images.
- Achieved 80.02% accuracy.
- Implemented dropout, early stopping, and data augmentation.
- 📄 Read Full Report (PDF)
🔹 Regression Analysis on Gene Expression
Gene Expression Regression Report
Explored regression methods for risk factor identification across thousands of genes.
- Used Lasso Regularization, PCA, and cross-validation to reduce dimensionality.
- Applied non-parametric smoothing (locopoly, B-spline, additive models).
- Evaluated models using R², residual diagnostics, and bootstrap CI.
🔹 Finance: Personal Loan Classification
Loan Classification Analysis
Classification model to predict which liability customers are likely to accept a personal loan offer.
- Techniques: decision trees, bagging, boosting, random forest.
- Reduced false negative rate from 0.13 to 0.08 using MLR3 threshold tuning.
- Identified variable correlations for client segmentation.
🔹 Unsupervised Learning: Generative Models
Generative Modeling & Clustering
Used Kernel Density Estimation and Gaussian Mixture Models to identify latent clusters in the data.
- Synthesized new data points from fitted distributions.
- Clustering via Expectation-Maximization (EM).
- Focused on High Density Regions.
🔹 Shakespeare Network & Text Analysis
Hamlet Character Network
Analyzed character relationships in Hamlet through dynamic network graphs.
- Built a graph from lines spoken per act.
- Applied community detection to trace plot evolution and key figures.
🧠 Skills Highlighted
- Data Cleaning & Preprocessing
- Supervised & Unsupervised Learning
- Dimensionality Reduction
- Deep Learning with CNNs
- Network & Text Mining
- Visualization & Reporting
Feel free to explore the notebooks and reports. If you’d like to collaborate or have questions, feel free to connect with me!