Implementation of the Naive Bayes Classifier for Sentiment Analysis of Shopee E-Commerce Application Review Data on the Google Play Store
Adilia Tri Rizkya,Rianto Rianto,Acep Irham Gufroni
DOI: https://doi.org/10.37058/jaisi.v1i1.8993
2023-11-21
Abstract:E-commerce in Indonesia is growing very quickly every year. The Ministry of Communication and Information (KEMKOMINFO) stated that Indonesia is the 10th largest e-commerce growth country with score 78%. One of the effects from increasing number of internet users in Indonesia is the mushrooming of shopping activities through internet media. This causes internet users want everything that instant and easy. Knowing this, most business people use it to market their products, especially in the field of goods and services. As it grows, e-commerce becomes easier to use and download. One example of an e-commerce application that is in great demand is Shopee and can be downloaded via the Google Play Store. Google Play Store has a review feature which contains user comments about the downloaded apps. Sentiment analysis is carried out to extract information related to Shopee E-commerce. The Naïve Bayes Classifier algorithm is suitable for use in sentiment analysis because this algorithm is purposeful as a classification method into positive and negative categories. The data was used from November 2022 to January 2023. From a total of 4902 review data obtained, after going through preprocessing, translation and then classification, the total data is obtained that is 4849 review data. From the data obtained it is classified 2348 positive reviews, 1259 neutral reviews, and 1242 negative reviews. Based on the results of the naive Bayes classifier method and testing with the confusion matrix, an accuracy value of 79% has been obtainednprecision 77%, recall 86%, and f1-score 81% on positive sentiment with support 2127. For neutral sentiment with an accuracy value of 83%, precision 87%, and recall 85% with support 1209, while for negative sentiment is with an accuracy value of 78%, precision 64%, and recall 70% with support 1513. From this data it is obtained micro AVG values for precision 80%, recall 79%, f1-score 79%, and support 4849, then for weighted average for precision 79%, recall 79%, f1- score 79%, and support 4849.