A machine learning model for fault prediction.
Sensor measurements are used to predict the failure of an expensive machine component. The sensor that best predicts component lifetime is identified. Then an XGBoost forecast model is fitted to the data.
The files work at least on Ubuntu 16.04 using Anaconda 4.5.12
The repository includes scripts for data extraction and analysis, feature selection, model fitting and prediction.
- data_extraction_and_analysis.ipynb is a Jupyter notebook for data extraction and analysis
- feature_selection_and_modeling.ipynb is a Jupyter notebook for feature selection, model fitting and prediction.
- expanding_window.py is a Python function for walk-forward cross-validation
- data includes data input and output files (remember to unzip sensor measurements)