Penerapan Algoritma K-Nearest Neighbors (KNN) Dalam Analisis Sentimen Ulasan Pengguna Aplikasi JMO (Jamsostek Mobile)

Abstract

JMO (Jamsostek Mobile) is an official application launched by BPJS Ketenagakerjaan, designed to support workers in facilitating access to social protection. However, several obstacles are experienced by application users, such as errors in the application, difficulties logging in, and JHT balances not being displayed. To obtain a general overview of JMO application user sentiment, an evaluation is needed to capture user concerns. Therefore, this study will explore the precision of the KNN algorithm in analyzing sentiment from JMO application user reviews. The objectives of this research are to identify and analyze user feedback, classify overall sentiments, implement sentiment analysis, and test the accuracy of the KNN algorithm. This study adopts a quantitative approach, applying numerical data analysis through text mining techniques. From a dataset of 10,000 reviews collected via web scraping and refined Preprocessing, 7,185 reviews were obtained, revealing that 51.38% expressed positive sentiment. The KNN algorithm achieved its highest accuracy 76.2%, precision 76.2%, recall 78.0%, and F1-score 77.1% at K = 21 under 90%-10% data split. Furthermore, the model’s AUC score of 0.7617 indicated fair and reasonably good performance. These findings suggest that the KNN classification model is capable of providing balanced classification results across classes, leading to a fairer and less biased evaluation.