Comparison of SVM, Naive Bayes, and Logistic Regression for LinkedIn Reviews Sentiment Analysis

The rapid development of digital technology has transformed the way people search for jobs, with LinkedIn emerging as the world’s largest professional social media platform. Many users express their opinions about the application through reviews on the Google Play Store, reflecting both positive and negative sentiments regarding their experiences. This study aims to conduct sentiment analysis on LinkedIn user reviews by comparing three classification algorithms: Support Vector Machine (SVM), Naïve Bayes, and Logistic Regression. The research process involves data collection, text preprocessing, feature extraction, and model evaluation using accuracy, precision, recall, and F1-score metrics. The results indicate that all three algorithms are capable of classifying sentiments effectively, with Logistic Regression achieving the best performance, obtaining an accuracy of 88.53%, a precision of 94% for negative reviews and 83% for positive reviews, as well as a recall of 84% for negative reviews and 94% for positive reviews. In comparison, SVM achieved an accuracy of 87.79%, while Naïve Bayes reached 83.44%. These findings highlight that Logistic Regression outperforms the other models in sentiment analysis of LinkedIn reviews, making it a reliable method for understanding user perceptions and supporting application improvement.

Keywords: Feature Extraction, Model Evaluation, Text Preprocessing, User Perceptions, User Review