The Effect of Data Splitting Ratio and Vectorizer Method on the Accuracy of the Support Vector Machine and Naïve Bayes Model to Perform Sentiment Analysis

The main information retrieved from public opinions can be understood by carrying out a sentiment analysis. The validity of the retrieved information depends on the performance of the algorithm used to perform the sentiment analysis., where sentiment analysis parameters themselves need to be considered. Thus, it is important to determine the best level parameter and algorithm in order to obtain a valid conclusion from a sentiment analysis. This study aims to determine the effect of the data splitting ratio (0.5, 0.6, 0.7), and vectorizer (CountVectorizer, TF-IDF) on the accuracy of the Support Vector Machine and Naïve Bayes model while performing sentiment analysis. The results of the research show that the data splitting ratio and vectorizer have a significant effect on the accuracy of the SVM and Naïve Bayes model performance. However, the interaction between the data splitting ratio and the vectorizer method does not significantly influence the accuracy of both models.