COMPARISON OF RANDOM FOREST, SUPPORT VECTOR MACHINE AND NAIVE BAYES ALGORITHMS TO ANALYZE SENTIMENT TOWARDS MENTAL HEALTH STIGMA

Abstract

Advances in technology, especially the internet, have significantly changed the way people communicate, including social media. Social media facilitates more effective and efficient online communication. Twitter has 18.45 million users in Indonesia by 2022. Discussion of mental health stigma on twitter, increased 17% in 2021 compared to the previous year. Lifestyle transformation, social pressures, and technological advancements have created new challenges in maintaining individual mental health. Discussions of mental health issues have become pros and cons on twitter. The tendency of twitter users in posting content can be known by means of sentiment analysis. Therefore, sentiment analysis can be used to classify comments and tweets related to mental health stigma into negative, positive and neutral. So, it is expected to provide a number of significant benefits in the aspect of managing mental health issues. The methods used to analyze sentiment towards mental health stigma are Random Forest, Support Vector Machine (SVM) and Naïve Bayes algorithms. Based on the research that has been done, it produces 3,095 data for the period 2020-2023. After preprocessing and labeling the data, 1,635 data (negative class), 633 data (positive class) and 208 data (neutral class) were obtained. The SVM model test results show an accuracy of 86.11%, the Random Forest model shows an accuracy of 82.55%, while the Naive Bayes model shows an accuracy of 78.19%. Therefore, it can be concluded that SVM has the best performance in classifying tweets containing mental health stigma.