A comprehensive comparison of three dimensionality reduction techniques — PCA, t-SNE, and UMAP — applied to the Fashion-MNIST dataset (3,000 samples). The analysis covers quantitative evaluation (silhouette scores, k-NN preservation, retrieval precision, confusion analysis), advanced experiments (cluster density, embedding stability, distance preservation), and real-world application analogies. Interactive scatter plots with hover-to-view-image functionality are included for exploratory visualization.

This advanced Market Basket Analysis on Instacart data benchmarks the classic Apriori algorithm against the high-efficiency FP-Growth method to ensure scalability. Moving beyond standard metrics, the project employs Kulczynski, Certainty Factor, and Added Value to isolate robust rules, while also implementing Sequential Pattern Mining (cSPADE) and multi-level hierarchy analysis. Supported by rigorous memory optimization and statistical testing, this study uncovers deep, actionable insights into consumer purchasing behaviors and organic product ecosystems.

This project explores movie consumption patterns using a hybrid unsupervised learning approach, moving beyond standard market basket analysis. I integrated User Clustering (K-Means) to first segment the audience, and then applied Sequential Pattern Mining (using the cSPADE algorithm) to uncover hidden viewing trajectories. The report also compares the efficiency of ECLAT vs. Apriori and uses interactive network visualizations to map out the cultural connections between film genres.