COMPARISON OF MACHINE LEARNING ALGORITHMS FOR SENTIMENT ANALYSIS OF DIGITAL IDENTITY APPLICATION USERS

Abstract

In the rapidly evolving digital era, the Population Identity Application (IKD) plays a crucial role in streamlining civil administration processes in Indonesia, allowing easier and faster access to population services. This study aims to explore the application of machine learning algorithms in analyzing user responses to the IKD application. Three popular algorithms: Support Vector Machine (SVM), K-Nearest Neighbors (K-NN), and Naïve Bayes were chosen to classify sentiment from 1301 user reviews on the Google Play Store into positive and negative categories. After performing data preprocessing such as tokenization and stemming, hyperparameter optimization was conducted using GridSearchCV to enhance classification accuracy. The research results indicate that the SVM algorithm, optimized with hyperparameters, including the use of the rbf kernel and a parameter value of C = 1, achieved the highest accuracy of 85.60%, making it the most effective method for sentiment classification of the IKD application. These findings provide valuable insights for the government and developers in refining the features and performance of IKD, contributing to the efficiency and security of digital administration in Indonesia. Furthermore, this study opens opportunities for further development that is more responsive to user needs and expectations in the future.

Keywords

• digital identity application
• k-nearest neighbors
• naïve bayes
• sentiment analysis
• support vector machine