Pirelli - HIT Industrial AI Challenge 2023

This code repository contains all the scripts/notebooks used during the 12 weeks research based challenge. Using statistical inference and machine learning to analyse and forecast the curing machines anomalies.

Table of content

• Data
  • data per ipcode
  • data per machine
• Data generator
• Notebooks
• Prediction
• Authors

Data

Contains all the preprocessed data, divided by year.

• data per ipcode (type)
• data per machine

Data generator

Synthetic data generator

Notebooks

Contains all the data exploration notebooks, in particular:

• Data_Exploration_year_2022.ipynb: raw data exploration
• Machine_exploration.ipynb: contains the data exploration of the preprocessed data per machine, most relvant for the model creation.
• IPcode_exploration: outliers detection and analysis
• Others: contains the data exploration of the preprocessed data per year, not relevant for the model creation.

Prediction

Contains all the scripts for the model creation and the prediction. The most relevant are:

• machine_dataset_values_generator.py: generate the preprocessed data per machine
• evaluation.py: contains the models used and tested, the data preprocessing, tranformation and fitting to the models, model training, evaluation metrics, predictions, plots and model explaination with lime.

Authors

• Anna Elise Høfde
• Elia Zonta
• Erik Nielsen
• Usama Zafar