Optimasi Algoritma Support Vector Machine Berbasis PSO Dan Seleksi Fitur Information Gain Pada Analisis Sentimen

Abstract

Sentiment analysis is a method for processing consumer reviews. This study examines the application of the Support Vector Machine (SVM) algorithm based on PSO and Information Gain as feature selection to filter attributes as a form of optimization. Algorithm implementation in sentiment analysis is carried out by applying a test scenario to measure the level of accuracy of the several parameters used. Selection of the Information Gain feature using the top-k parameter yields an accuracy value of 85.3%. Algortima optimization applying information gain feature selection on the PSO-based SVM resulted in an optimal accuracy rate of 86.81%. The resulting increase in accuracy is 18.84% compared to the application of classic SVM without PSO-based information gain feature selection. Applying information gain feature selection on the PSO-based SVM algorithm can increase the accuracy value in the online sentiment review analysis.