Tech Stack: Java, Spring Boot, ELKI (Environment for Developing KDD-Applications Supported by Index-Structures)
🧠 Project Overview This project focuses on anomaly detection in Breast Cancer data using the Local Outlier Factor (LOF) algorithm with K-Nearest Neighbors (KNN) and High Contrast Subspaces (HICS) for effective subspace selection.
⚙️ Key Features ✅ LOF-based Anomaly Detection: Applied the LOF algorithm using KNN to detect anomalies in medical data.
✅ High Contrast Subspace Selection: Implemented HICS to identify relevant subspaces for better feature contrast and anomaly identification.
✅ Custom Feature Selection Methods:
selectHighContrastSubspaces
selectRandomFeatures
filterData
calculateContrast
✅ Outlier Reporting: Identified and printed the top 10 outliers based on LOF scores.
✅ Model Evaluation:
Evaluated results using a Confusion Matrix
Plotted the ROC curve and computed the AUC to measure detection performance.
This project demonstrates a practical application of subspace analysis and anomaly detection in medical datasets using Java and ELKI.