PEMODELAN PREDIKSI KINERJA SISWA DENGAN HYBRID PARTICLE SWARM OPTIMIZATION-SUPPORT VECTOR REGRESSION (PSO-SVR) DAN GENETIC ALGORITHM-SUPPORT VECTOR REGRESSION (GA-SVR)
Total View This Week0
Total View This Week0
Institusion
Universitas Sriwijaya
Author
APRIYADI, M. RIKI
Ermatita, Ermatita
Rini, Dian Palupi
Ermatita, Ermatita
Rini, Dian Palupi
Subject
Q325.5 Machine learning
Datestamp
2023-05-10 02:08:59
Abstract :
Improving the quality of education is very important in the development of a country. Identifying representative educational data and conducting predictive analysis of student performance in an educational institution is very important to assist management in determining strategies and making decisions to improve student performance. This study modeled student performance using Hybrid Particle Swarm Optimization-Support Vector Regression (PSO-SVR) and Genetic Algorithm-Support Vector Regression (GA-SVR). This study uses predictive modeling of student performance with SVR modeling. PSO and GA are used as feature selections to identify relevant factors that influence student performance. The feature selection results, which are selected features, are then predicted with SVR to produce FSPSO-SVR and FSGA-SVR modeling. This study also performs hyperparameter optimization on SVR to obtain the best hyperparameters using PSO and GA, resulting in PSO-SVR (PSVR) and GA-SVR (GSVR) modeling. In the final stage, a comparison of the results of regression performance metrics was carried out from this study and previous studies. From the results of this study, the SVR modeling produced an RMSE value of 4.519; the FSPSO-SVR modeling produced an RMSE value of 2.954; the FSGA-SVR modeling produced an RMSE value of 4.447; the PSVR modeling resulted in an RMSE value of 1.608; the GSVR modeling had an RMSE value of 1.830. Meanwhile, from previous research conducted by Cortez on SVR modeling, the RMSE value was 2.090. The comparison of regression performance metrics found that the best student performance prediction modeling used PSVR modeling with a decrease in error values of 64.42% compared to SVR modeling in this study and a decrease in error values of 23.06% compared to previous studies.
Improving the quality of education is very important in the development of a country. Identifying representative educational data and conducting predictive analysis of student performance in an educational institution is very important to assist management in determining strategies and making decisions to improve student performance. This study modeled student performance using Hybrid Particle Swarm Optimization-Support Vector Regression (PSO-SVR) and Genetic Algorithm-Support Vector Regression (GA-SVR). This study uses predictive modeling of student performance with SVR modeling. PSO and GA are used as feature selections to identify relevant factors that influence student performance. The feature selection results, which are selected features, are then predicted with SVR to produce FSPSO-SVR and FSGA-SVR modeling. This study also performs hyperparameter optimization on SVR to obtain the best hyperparameters using PSO and GA, resulting in PSO-SVR (PSVR) and GA-SVR (GSVR) modeling. In the final stage, a comparison of the results of regression performance metrics was carried out from this study and previous studies. From the results of this study, the SVR modeling produced an RMSE value of 4.519; the FSPSO-SVR modeling produced an RMSE value of 2.954; the FSGA-SVR modeling produced an RMSE value of 4.447; the PSVR modeling resulted in an RMSE value of 1.608; the GSVR modeling had an RMSE value of 1.830. Meanwhile, from previous research conducted by Cortez on SVR modeling, the RMSE value was 2.090. The comparison of regression performance metrics found that the best student performance prediction modeling used PSVR modeling with a decrease in error values of 64.42% compared to SVR modeling in this study and a decrease in error values of 23.06% compared to previous studies.
File :
RAMA_21001_03043681823001.pdf
RAMA_21001_03043681823001_TURNITIN.pdf
RAMA_21001_03043681823001_0013096707_0023027804_01_front_ref.pdf
RAMA_21001_03043681823001_0013096707_0023027804_02.pdf
RAMA_21001_03043681823001_0013096707_0023027804_03.pdf
RAMA_21001_03043681823001_0013096707_0023027804_04.pdf
RAMA_21001_03043681823001_0013096707_0023027804_05.pdf
RAMA_21001_03043681823001_0013096707_0023027804_06_ref.pdf
RAMA_21001_03043681823001_0013096707_0023027804_07_lamp.pdf
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
