This repository contains the complete codebase accompanying the paper:

A Survival-Theoretic Interpretation of Change Point Detection under Autocorrelation and Transient Deviations
Presented at the 8th International Conference on Complex Dynamical Systems and Applications (CDSA 2026)
Dhirubhai Ambani University, Gandhinagar, India.

The repository benchmarks standard change point detection (CPD) algorithms under autocorrelated time series and evaluates them using a survival-theoretic risk metric combining false alarms and detection delay.

• PELT (L2)
• Binary Segmentation
• Segment Neighborhood
• Robust RBF
• AR(1) + PELT
• Bayesian Offline CPD (diagnostic)

• main.py – data loading and preprocessing
• cpd.py – CPD method execution
• pelt.py, rbf.py, ar1_pelt.py – sensitivity diagnostics
• evaluation.py – false alarm, delay, MRL, and risk computation
• run_eval.py – runs full evaluation pipeline
• plot_results.py – visualization of detections and delays

python main.py
python cpd.py
python run_eval.py