MEMPREDIKSI PRODUCTIVITY SCORE BERDASARKAN HOURS PER WEEK

Abstract

Individual productivity is an important indicator in assessing performance both in the work and educational environment. This research aims to predict productivity scores based on the hours per week variable, namely the number of hours a person spends working per week. Using several approaches using regression methods, we explore the relationship between the number of hours worked per week and productivity levels. The dataset used in this research was taken from Kaggle. The best prediction accuracy is found in the SVR RBF algorithm using training data and test data of 70%:30% and this data is most suitable using SVR RBF using training data and test data of 70%:30%. This research provides insight into the optimal limits of working hours that can support productivity. The best MAE value is using the Linear SVR algorithm with training data and test data of 70%:30% of 8.801. A good RMSE value using the SVR RBF algorithm with training data and test data of 70%:30% is 11.391. The best R2 value is found in SVR RBF with training data and test data of 70%:30% of 0.062. The worst performance, worst prediction accuracy, and least suitable data use the SVR RBF algorithm with training data and test data 50%:50%. The MAE value that has poor performance is the SVR RBF algorithm with training data and test data of 50%:50% of 10.286. The RMSE value which has poor performance uses the SVR RBF algorithm with training data and test data of 50%:50% with a value of 13.046. The R2 value that has poor performance is the SVR RBF algorithm with training data and test data of 50%:50% with a value of 0.017.