Performance Analysis of the K-Nearest Neighbors (K-NN) for Sentiment Analysis of Online Loan Application X

Abstract

The digital economy in Indonesia is growing rapidly, including in online lending services. Application X is one of the popular online lending applications, offering users convenience in applying for loans online. This research employs sentiment analysis on user reviews of Application X to understand their preferences and needs. The K-Nearest Neighbours (K-NN) method is applied as the primary algorithm for sentiment classification. Data collected through user review scraping undergoes a series of preprocessing stages, such as tokenization, stop word removal, and stemming, aimed at improving data quality. The K-NN model is tested in various scenarios to achieve the best results. The best scenario reveals that the highest accuracy is achieved by the K-NN model when the stop word removal process is not applied during the data preprocessing stage where the accuracy without using the stop word process was 92.9%, compared to 89.9% when using stop words.