<![CDATA[This study aims to classify user opinions on IMDb movie reviews using the Multinomial Naïve Bayes algorithm. The dataset consists of 50,000 reviews, evenly distributed between 25,000 positive and 25,000 negative reviews. The preprocessing stage includes cleaning, case folding, stopword removal, tokenization, and lemmatization using the NLTK library. Text features are represented through the TF-IDF method to capture the significance of each word in the documents. The Multinomial Naïve Bayes model was trained using the hold-out validation technique with an 80:20 split for training and testing data. Hyperparameter tuning of α (Laplace smoothing) was conducted to enhance model stability and accuracy. The model’s performance was evaluated using accuracy, precision, recall, and F1-score metrics, supported by a confusion matrix visualization. The results show that the model achieved an accuracy of 87%, with precision of 87.9%, recall of 85.4%, and an F1-score of 86.6%. In comparison, Logistic Regression as a baseline algorithm achieved an accuracy of 91%. Nevertheless, the Naïve Bayes algorithm remains competitive and computationally efficient for large-scale text data, making it highly relevant for sentiment analysis of movie reviews.]]>
Copyrights © 2025
