Perbandingan Evaluasi Kernel Support Vector Machine dalam Analisis Sentimen Chatbot AI pada Ulasan Google Play Store

Celine Mutiara Putri (Universitas Islam Negeri Sultan Syarif Kasim Riau)
M. Afdal (Universitas Islam Negeri Sultan Syarif Kasim Riau)
Rice Novita (Universitas Islam Negeri Sultan Syarif Kasim Riau)
Mustakim Mustakim (Universitas Islam Negeri Sultan Syarif Kasim Riau)

Jurnal Teknologi Sistem Informasi dan Aplikasi
Vol. 7 No. 3 (2024): Jurnal Teknologi Sistem Informasi dan Aplikasi

AI (Artificial Intelligence) is becoming very important these days due to its ability as a personal assistant to increase efficiency, automate routine tasks, and speed up manual processes. AI chatbot are one of the practical applications of AI in language understanding, have various benefits and drawbacks that cause various comments from users in the review column on the Google Play Store. This research discusses sentiment analysis of AI chatbot application reviews using four SVM kernels. Labeling uses InSet Lexicon and hyperparameters to produce the best parameters. The purpose of the research is to find out how users respond to interactions with ChatGPT, Perplexity AI, and Bing Chat and prove whether the kernel in SVM can increase the accuracy value. The percentage division between test data and training data is 70:30, 80:20, and 90:10, data labeling using 2 sentiment classes and 3 sentiment classes, and using and not using the SMOTE Oversampling technique. The experimental results obtained the highest accuracy using SVM kernel Linear scenario 90:10 with an accuracy value of 92.68%.