ENHANCING SENTIMENT ANALYSIS ACCURACY IN DELIVERY SERVICES THROUGH SYNTHETIC MINORITY OVERSAMPLING TECHNIQUE
DOI:
https://doi.org/10.33480/jitk.v9i2.5058Keywords:
random forest, sentiment analysis, smote, SVMAbstract
Paxel is one of the delivery services that use the application. On Google Play, there are more than 10 thousand users leaving reviews. From this review data, a sentiment analysis was then carried out to determine the level of user satisfaction with Paxel's services. The methods used in this study are Random Forest (RF) and Support Vector Machine (SVM), as well as applying Synthetic Minority Oversampling Technique (SMOTE) to overcome data imbalance. The results showed that the method testing by dividing the data into two, namely training data and testing data by 80:20, stated that by applying the SMOTE, a higher accuracy value was obtained, where the accuracy of the RF method reached 91%, and the SVM method reached 87%. The level of user satisfaction with Paxel services tends to be neutral. This can be seen in the classification of the RF method with F1-Score values for the Positive class 89%, Neutral class 93%, and Negative class 92%.