ETD

• Boruta
• churn
• customer churn
• feature selection
• modello di previsione
• prediction model
• telco
• telecommunications

The goal of the thesis is the evaluation of the impact of the feature selection process on the performance of a customer churn prediction model in the telecommunications sector. The feature selection technique used in the paper is a wrapper algorithm Boruta. Various experiments were performed with changes in the parameter values of the Boruta algorithm and a final set of most important features was selected that was used for the prediction model. The study shows the importance of the process of feature selection and it’s impact on the average lift as a measure of performance of the prediction model.