Overview

This data mining project compares different clustering algorithms to analyze COVID-19 transmission patterns. By applying spectral clustering, K-means, and KNN clustering techniques, the analysis identifies relationships between peaks in transmission across different regions.

Key Objectives

• Algorithm Comparison: Evaluate performance of different clustering methods
• Pattern Discovery: Identify clusters of similar transmission patterns
• Peak Analysis: Understand the timing and relationships between transmission peaks
• Methodology Insights: Determine which clustering approach works best for epidemiological data

Methodology

• Data Collection: COVID-19 case data from multiple regions over time
• Preprocessing: Time series normalization and feature extraction
• Clustering Algorithms: Implementation of K-means, KNN, and Spectral Clustering
• Evaluation: Silhouette scores and cluster quality metrics

Clustering Approaches

• K-means Clustering: Traditional centroid-based clustering for pattern grouping
• KNN Clustering: Density-based approach using nearest neighbor relationships
• Spectral Clustering: Graph-based method capturing complex cluster structures

Technologies Used

• Python: Primary programming language
• Scikit-learn: Clustering algorithm implementations
• NumPy & Pandas: Data manipulation
• Visualization: Cluster visualization and comparison charts

Key Findings

Spectral clustering proved most effective at capturing non-linear relationships between transmission patterns, revealing geographic and temporal clusters that traditional methods missed.