Abstract

The objective of this research was to forecast the resignation of skilled technicians at a large automobile service center in the country using machine learning techniques. This study used the Random Forest algorithm along with the SMOTE (Synthetic Minority Oversampling Technique) method developed in Python. The research was conducted by preparing a questionnaire, with questions divided into 3 areas: personal factors; push factors and pull factors. Each question consisted of 31 subtopics. The total number of employees who responded to the questionnaire was 244 people, with 227 of them who were still working and 17 having resigned. Therefore, the data was unbalanced, requiring the creation of synthetic data using the Random Forest algorithm with the SMOTE technique in order to balance the two types of data. The experimental results showed that the model used to predict employee resignation using 3 types of input factors, personal factors + push factors, personal factors + pull factors, and personal factors + push factors + pull factors, was effective. When using personal factors and with only 2 push factors, it was found that the efficiency of the forecasting model had an accuracy of 100%, sensitivity of 100%, precision of 100% and F1-score of 100%. The results of the research showed that using Random Forest and SMOTE to address the data asymmetry problem resulted in high accuracy in the model's prediction performance. 

Statistics