Comparative Analysis of Classification Models for Competencies Certification: A Data Mining Approach

This study elaborates the application of various machine learning (ML) algorithm models to identify the classification of employee competency to take the competency certification needed to support the work. The goal is to choose the best ML model to improve accuracy, scalability, and fairness. The algorithms to be tested in this study are Logistic Regression, SVM, KNN, and Naive Bayes as traditional algorithms with Random Forest as an ensemble algorithm. This study took data from 13,830 employees who had been submitted in competency certification activities in 2024 at PT PLN (Persero). All models were measured through cross-validation on parameters such as accuracy, precision, recall, and F-1 score using the Python programming language on Jupyter Notebook. The best performing model on the F-1 score parameter was Logistic Regression which achieved the highest score of 0.9559. Meanwhile, Random Forest is the best model on the Precision parameter which is very important to identify employees who are truly incompetent to avoid losses due to human error. Based on this research, Logistic Regression and Random Forest can be prioritized to improve the accuracy of employee competency status in order to produce truly competent employees to strengthen operational performance and increase company revenue.