ANALISIS OPINI FILM PADA NETFLIX DENGAN ALGORITMA NAÏVE BAYES DAN SUPPORT VECTOR MACHINE MENGGUNAKAN SELEKSI FITUR CHI-SQUARE

Abstract

This research aims to analyse user opinions on films on the Netflix platform using the Naïve Bayes algorithm and Support Vector Machine. The focus of the research is to increase classification accuracy through feature selection using the Chi-square method. The data used is obtained through a web scraping process of user reviews on Google Play Store. Automatic labeling is supported by the Transformers library, resulting in 131 positive labels and 869 negative labels from 1000 reviews. The research stages include data crawling, automatic labeling using the Transformers library, pre-processing (case folding, tokenisation, stopword removal, normalisation, and stemming), weighting with the TF-IDF method, and testing model accuracy using data split ratios of 90:10, 80:20, and 70:30. The findings of the study indicate that the Support Vector Machine algorithm reached an accuracy rate of 92.5% using the 80:20 data split, whereas its Chi-square enhanced variant attained 91.5% accuracy on the same dataset. Meanwhile, the Naïve Bayes classifier recorded an accuracy of 82%, and its Chi-square integrated version yielded 79%. These results suggest that incorporating Chi-square did not enhance the predictive performance of either the Naïve Bayes or Support Vector Machine approaches in this research.