Optimizing Genetic Algorithms for Sentiment Analysis of Apple Product Reviews Using SVM

Abstract

Online reviews have the potential to provide buyers with insights about products such as quality, performance and recommendations. Website is one of the media that contains information or reviews provided by individuals, groups or organizations about an object or topic, one of which is Apple products. This study analyzes consumer sentiment reviews of Apple product users consisting of 200 reviews which will be classified into positive opinions and negative opinions using the Support Vector Machine algorithm and the application of genetic algorithms (GA) to obtain optimal accuracy values. The stages of this research are, firstly collecting a dataset, the second is preprocessing data. Third, the sentiment analysis process uses SVM and GA as optimization techniques. Fourth, do the validation process on the accuracy results obtained using the Confusion Matrix and ROC Curve. The results of this study indicate that Apple product review sentiment analysis produces the best accuracy of 70.00% and AUC 0.924 for SVM algorithm. Whereas the SVM + GA algorithm produces 85.76% accuracy and AUC 0.945, so that the accuracy value increases by 15.76% and the AUC 0.021 on the SVM model when compared before optimization with genetic algorithms (GA) is performed