APLIKASI KLASIFIKASI SENTIMEN PADA ULASAN SMARTPHONE DI SITUS JUAL BELI ONLINE BERBASIS WEB MENGGUNAKAN NAIVE BAYES DENGAN TF-IDF

Abstract

Consumer satisfaction can be seen through the product review column and everyone is free to see it, but the problem that occurs is that limited human resources make business owners and prospective smartphone product buyers unable to read all the reviews in a short time. For that we need a method that can assess these reviews automatically. The solution to overcome the limitations in assessing review data is to use machine learning because it produces analysis with a statistical basis with a fairly reliable algorithm with high accuracy that can be used to classify sentiment analysis on smartphone product reviews called Naive Bayes. Web-based classification applications using Naïve Bayes can be made through 5 stages of SDLC written into source code using the web programming language on the Laravel Framework. Before doing the classification, each word in the review data was first given a weight using TF-IDF and all words were successfully weighted and then stored in the database. The results of testing the accuracy of the application achieve very good results, namely 93% and the results of precision and recall which reach numbers above 85% indicate that the Naive Bayes algorithm is functioning well in the application.