This is the code for my bachelor thesis which was under the title "Prediction of depressive episodes using biomedical sensors with machine learning techniques".
The behaviours of patients with depression are usually diffcult to predict because the patients demonstrate the symptoms of a depressive episode without warning at unexpected times. The goal of my thesis was to examine different machine learning approaches to detect such times and predict possible depressive episodes. The work also comprised a preprocessing pipeline for transforming and organizing the input data. The final assessment indicates current and future application possibilities of machine learning in the depression detection context.
Tools/frameworks: Keras, tf, scikit-learn.
The code is divided into four major parts:
- python files that are responsible for the data preprocessing parts.
- files for the NN work done
- files for the Random Forest implementations.
- files for the SVM implementaions.
apart from the general structure, the code is throughly documented.