Challenge: A large fleet of devices requires maintenance to prevent device failure. This repository presents a predictive analysis to help with device maintenance by predicting failure given a series of device attributes, measured daily over the course of 11 months in 2015. The final analysis employs a neural network model with a F1 Score of .96 and a ROC-AUC score of .86.
Dataset: 124,164 daily readings from 1163 devices across 9 attributes related to device failure.
- 1. Data cleaning
- 2. EDA by device ID
- 3. Trimming the dataset
- 4. EDA by other variables
- 5. Feature engineering
- 6. Logistic Regression Results
- 7. Neural Network Results
Survival Analysis
https://scikit-survival.readthedocs.io https://www.kdnuggets.com/2017/11/survival-analysis-business-analytics.html https://www.kdnuggets.com/2020/07/guide-survival-analysis-python-part-3.html https://humboldt-wi.github.io/blog/research/information_systems_1920/group2_survivalanalysis/