Repository of the experimental setup for the paper "Detection of Component Degradation: A Study on Autoencoder-based Approaches". It should be noted that the following packages are needed to run our experimental setup:
- pyNeVer and all its required packages
- sklearn
- pandas
- matplotlib
Furthermore, given the size of the data, the user needs to autonomously download the dataset and paste the folder "oneyeardata/" in the folder data/.
The results, models and graph presented in the paper can be found in the folder exp_paper_res/. If the user want to re-generate the models, the following scripts must be launched in order:
- data_preprocessing.py
- models_generation.py
- loss_analysis.py
The results obtained by this process may slightly differ from the one presented in the paper due to the pseudo-randomic nature of some algorithms used.