Utilization of machine learning for Parkinson's disease prediction

The bachelor thesis focuses on the use of machine learning for the prediction of Parkinson´s disease, specifically utilizing speech biomarkers. The theoretical part of the thesis provides a brief introduction to Parkinson´s disease and its impact on speech in the early stages of the disease. Subsequently, the decision tree and random forest classification models, along with evaluation metrics and methodology, are presented. In the practical part, a relevant dataset is first selected, explored and adjusted. Next, a suitable prediction algorithm is chosen, in this case random forest, and applied to the adjusted dataset. The imbalance of the dataset is resolved with 10-fold cross-validation. Furthermore, hyperparameter optimization is performed. Finally, the final model is built and evaluated. The practical part is accompanied by a correlation analysis that describes the relationship between the attributes of the adjusted dataset and an analysis of the importance of individual predictors in the final model. The thesis demonstrates the potential use of the random forest algorithm for predicting Parkinson´s disease on real data and provides an R script to replicate the approach and results.