Rivana Dwi Cahyani, Putri Taqwa Prasetyaningrum

Sentiment Analysis of User Reviews for AI Applications: Evaluating SVM, Logistic Regression, and Random Forest

Introduction

Sentiment analysis of user reviews for ai applications: evaluating svm, logistic regression, and random forest. Evaluate SVM, Logistic Regression, and Random Forest for sentiment analysis of Indonesian AI app user reviews. Discover Random Forest's 99.62% accuracy for developer insights.

Abstract

The rapid growth of AI applications such as CICI, GROK, and Gemini has resulted in a large volume of user reviews on platforms like the Google Play Store, making sentiment analysis a critical tool for understanding user perceptions. This study compares the performance of three machine learning models: Random Forest, Support Vector Machine (SVM), and Logistic Regression in classifying sentiments in 3,500 Indonesian-language reviews. A hybrid feature extraction approach, combining sentiment lexicons with TF-IDF, was applied to improve sentiment classification accuracy. The models were evaluated based on accuracy, precision, recall, and F1-score. Results indicated that all models achieved an accuracy greater than 96%, with Random Forest providing the most consistent and accurate results, achieving an overall accuracy of 99.62%. While SVM excelled in classifying positive and negative sentiments, it faced challenges with neutral reviews due to the ambiguity and overlap in sentiment expression. Logistic Regression also showed strong performance, especially on structured reviews. The findings suggest that Random Forest is the most robust and reliable model for sentiment analysis, particularly in handling diverse AI application reviews. These results offer practical insights for developers seeking to improve application performance by leveraging sentiment analysis on user feedback.

Review

This study, "Sentiment Analysis of User Reviews for AI Applications: Evaluating SVM, Logistic Regression, and Random Forest," tackles a highly pertinent subject given the rapid proliferation of AI applications and the subsequent demand for understanding user perceptions through feedback. The authors' objective to compare the performance of Random Forest, SVM, and Logistic Regression models on a substantial dataset of 3,500 Indonesian-language reviews is commendable, offering a valuable contribution to cross-lingual sentiment analysis. A notable strength of the methodology lies in the hybrid feature extraction approach, combining sentiment lexicons with TF-IDF, which appears to be a key factor in achieving high classification accuracies. The research presents compelling results, with all evaluated models demonstrating impressive accuracy exceeding 96%. Random Forest particularly stands out as the most consistent and accurate performer, achieving an exceptional overall accuracy of 99.62%, making a strong case for its suitability in this application domain. The abstract also provides nuanced insights into the performance of the other models, noting SVM's struggles with neutral reviews due to inherent ambiguity, and Logistic Regression's strong performance on structured feedback. These detailed findings are crucial for developers, offering practical guidance on selecting appropriate models based on the specific characteristics of the review data. While the reported accuracies are exceptionally high and indicate a robust methodology, future work could benefit from a more in-depth discussion on the generalizability of these findings beyond the Indonesian language context or the specific AI applications reviewed. A deeper analysis into the specific factors contributing to Random Forest's superior performance and a more elaborate breakdown of the challenges SVM faced with neutral reviews would further enrich the study's methodological contributions. Nonetheless, this paper delivers valuable insights, reinforcing the efficacy of machine learning in processing user feedback for AI applications and providing actionable recommendations for developers to enhance application performance.

Full Text

Comments

You need to be logged in to post a comment.