RANCANG BANGUN APLIKASI ANALISIS SENTIMEN ULASAN APLIKASI DI GOOGLE PLAY STORE MENGGUNAKAN METODE LOGISTIC REGRESSION DAN LEXICON-BASED APPROACH

Abstract

In the increasingly advanced digital era, user reviews of mobile applications on platforms such as the Google Play Store have become an important source of information for both developers and users. These reviews can reflect user satisfaction, complaints, and expectations for an application. However, due to the ever-increasing volume of reviews, it is very difficult for developers to read and evaluate all opinions manually. On the other hand, available reviews often do not reflect the actual conditions due to the presence of fake, biased, or manipulative reviews. Therefore, an automated system is needed that can accurately group reviews based on sentiment polarity. This study aims to design and build a web-based application using the Streamlit framework that can perform sentiment analysis on application reviews on the Google Play Store. The method used is a combination of Logistic Regression and Lexicon-Based Approach with text preprocessing stages such as cleansing, tokenizing, stemming, and transformation using TF-IDF. This study uses the CRISP-DM approach in building the application model and system. As part of the system design, a prototype model was created in the form of a use case diagram, use case scenario, and activity diagram to describe the functional requirements and the flow of user interaction with the system. The final result of this research is a sentiment analysis application capable of classifying reviews into positive and negative categories, as well as displaying the model evaluation results in the form of metric visualizations and word clouds. From the Dana application case study, based on the results of the model performance evaluation using the confusion matrix metric, the Logistic Regression model built achieved an accuracy of 91.2%, a precision of 91.73%, a recall of 98.46%, and an F1-score of 64.78%.