Predictive Maintenance using Automated Machine Learning

Predicting Air Pressure System (APS) failures in Scania trucks using a TPOT AutoML pipeline

This classification task is based on the Industrial Challenge 2016 at The 15th International Symposium on Intelligent Data Analysis (IDA). It was published in the UCI Machine Learning Library and is available on Kaggle.

Results of the original challenge:

Top 3 contestants	Score	Type 1 faults	Type 2 faults
Camila F. Costa and Mario A. Nascimento	9920	542	9
Christopher Gondek, Daniel Hafner and Oliver R. Sampson	10900	490	12
Sumeet Garnaik, Sushovan Das, Rama Syamala Sreepada and Bidyut Kr. Patra	11480	398	15

Automated machine learning (AutoML)

TPOT is a Python Automated Machine Learning tool that optimizes machine learning pipelines using genetic programming.

Usage

This repository contains

• a module for preprocessing the data set preprocessing.py.
• a script for executing the TPOT process
• a notebook for evaluating the results of the TPOT process
  • Python script
  • Jupyter notebook with results

Results

Score	Type 1 faults	Type 2 faults
11960	646	11