Sentiment Analysis of Game Review in Steam Platform using Random Forest

Steam provides a platform for buyers to write reviews of the software or games they have purchased. Developers will benefit from knowing the criticisms and suggestions given by their community. The number of reviews users give is so large that developers find it difficult to determine whether users like or dislike the games they create. In the Steam application, there is a rating system, but the ratings given by users do not always represent the content of the comments. Therefore, sentiment analysis is used to facilitate developers in understanding the sentiment of the reviews given by users. Sentiment analysis is used to solve this problem. In this research, the sentiment analysis method used is Random Forest with TF-IDF feature extraction in Bigram and Trigram. Based on the research results, scenario testing using Bigram TF-IDF instead of Trigram then in the preprocessing stage without Lemmatization achieved the highest performance. The average F1 score obtained was 62%.