Analisis Sentimen Ulasan Pelanggan Terhadap Tomoro Coffe Menggunakan Metode Random Forest Berbasis Text Mining

Abstract

Using a Random Forest technique based on text mining, this research seeks to examine consumer sentiment about Tomoro Coffee reviews. There were a total of 927 reviews culled from various online sources; 546 were favorable and 381 were unfavorable. Data collection, sentiment labeling, and text preparation (including cleaning, stopword removal, tokenizing, and stemming) were the study steps that came before data transformation using TF-IDF. After then, the dataset was divided into two halves, with training and testing data comprising 80:20 of the total. The Random Forest method was then used for classification. The accuracy was 80.65%, precision was 83.58%, recall was 70.10%, and the F1-score was 79.50%, according to the test findings. It was clear from the confusion matrix that the model excelled at detecting positive emotion as opposed to negative. Good, comfortable, and pleasant were the most often used positive emotion terms in the word cloud, while application, voucher, promotion, and payment were the most commonly used negative sentiment words. Based on the findings, it seems that Tomoro Coffee's product and service quality may be evaluated using the Random Forest approach, which is based on text mining. This method is effective for categorizing customer review sentiment.