Unsupervised-ML

1. 'Unsupervised MLs' folder contains:

Python code:

• 'unsupervised_classical_algorithms.py': Isolation Forest (IF), Local Outlier Factor (LOF), One-Class Support Vector Machine (OCSVM)
• mlp_ae_anomalies_detection.py': Creating the Multilayer Perceptron Autoencoder (create model, save model, load model, detect anomalies)
• 'conv_ae_anomalies_detection.py': Creating the Convolutional Autoencoder (create model, save model, load model, detect anomalies)

1. Next, there are 3 folders called 'Scenario 1', 'Scenario 2' and 'Scenario 3'. Each scenario refers to a distinct test dataset (Test Dataset 1, Test Dataset 2 and Test Dataset 3), which is used to test all 5 unsupervised machine learning algorithms (3 classical, 2 autoencoders).

For example: Scenario *:

• 'Test Dataset *' folder: .csv files with 'normal' and anomalies: aeration valve, sensor drift, sensor bias, sensor spike and sensor Precision Degradation (PD).
• 'Test-Classical-Unsupervised-ML' folder: contains the Python code for testing Isolation Forest (IF), Local Outlier Factor (LOF), One-Class Support Vector Machine (OCSVM) with Test Dataset *.

-'Test-Conv-AE' folder: contains the Python code for testing the Convolutional Autoencoder with Test Dataset * + Detection system (detection time for each fault)

-'Test-MLP-AE' folder: contains the Python code for testing the MLP Autoencoder with test Dataset *