Accuracy Rate of Least Square Support Vector Machine Method and Its Various Modifications: A Forecasting Evaluation on Multi-Type Data

Abstract

This research aims to compare the accuracy levels of the Least Square Support Vector Machine (LS-SVM) method and its modification with other algorithms in predicting various types of data. A quantitative approach with meta-analysis was employed, and the data were analyzed using JASP software, focusing on Mean Absolute Percentage Error (MAPE), Effect Size (ES) values, and Summary Effect (SE). The data analysis concludes that, overall, the LS method exhibits an accuracy rate of 92.7%, categorized as high, with an estimated coefficient value of 0.073. Based on the algorithm used, the analysis results with the LS method achieved an accuracy rate of 87.5%. The LS-SVM method demonstrates a higher accuracy level, reaching 95.4%, while the LS-Combination method attains the highest accuracy rate, namely 95.6%. In data classification, the analysis results indicate the highest accuracy level in economic and trade data, amounting to 95.6%. For social and demographic data, the coefficient value is 0.122 with an accuracy rate of 87.8%. Finally, in agricultural and mining data, the generated accuracy rate is 86.6%. These findings provide valuable insights into the performance of the LS method and its modifications with other algorithms in the context of forecasting various types of data.