Evaluating machine learning algorithms for sentiment analysis: a comparative study to support data-driven decision making

This research investigates the accuracy and robustness of sentiment analysis models through a comparative analysis of three distinct machine learning algorithms: Bernoulli Naive Bayes, Linear Support Vector Machines, and Logistic Regression. The primary objective is to assess the performance of these models across various domains and datasets in sentiment analysis tasks. The study employs data from the IMDb 500k movie reviews dataset, utilizing machine learning techniques for sentiment classification. Specifically, the selected algorithms—Bernoulli Naive Bayes, Linear Support Vector Machines, and Logistic Regression—are employed to train the dataset. Upon evaluating the models, the findings reveal notable differences in accuracy. Both LinearSVM and Bernoulli Naive Bayes achieved the highest accuracy, with each recording 89% when rounded to the nearest hundredth. However, LinearSVM slightly outperforms Bernoulli Naive Bayes in other performance metrics. In contrast, Logistic Regression records the lowest accuracy among the three algorithms. These results highlight the significance of algorithm choice in sentiment analysis tasks, with LinearSVM and Bernoulli Naive Bayes outperforming Logistic Regression. The research contributes valuable insights into the comparative performance of these algorithms, providing guidance for practitioners and researchers in choosing effective models for sentiment analysis across diverse datasets and domains.