Fundamental loss function for change point detection
Final prject, ML course by prof. Burnaev, prof. Zaycev, Skoltech, 2021
by "Prove you are not a robot"
Time series analysis plays important role in various fields including medicine, industry, finance, meteorology, etc. Change point detection(CPD) allows to detect changes in systems behaviour and understand the underlying structure of processes. In this project we propose a model for CPD with a customized loss function and explore its properties in experiments on synthetic and real-life time series.
Content:
Data - data for experiments
project.ipynb- notebook with project, includes data generation/selection, XGBoost's loss customization, experiments