/structural-break-detection

Implementation of the genetic algorithm for structural break detection in time series that chooses a piecewise autoregressive model using minimum description length principle

Primary LanguagePython

structural-break-detection

This is an implementation of a genetic algorithm for the final project for Machine Learning Workshop at the Warsaw University of Technology. The program analyzes the input from the phone's gyroscope and accelerometer (with React Native app, in the sensor-app repository) during the sleep. It is done by detecting structural breaks in time series with a genetic algorithm and minimum description length as objective function.

Exemplary determination of the breakpoints:

ml-ga