Hybrid Sentiment Analysis: Integrating Multinomial Naïve Bayes and Random Forest for Enhanced Text Classification

In this study, used ML techniques to analyse the sentiment of IMDB movie reviews. Before tokenization, stemming, and stopword removal, data preparation first required denoising text by eliminating HTML elements, special characters, and square brackets. Bag of Words (BoW) and TF- IDF vectorization were used to modify normalised data and capture word representations. To maximise vectorization, the computed document frequency (DF) and filtered words under predetermined limits (5–2000). Using TF-IDF representations, a number of classifiers were constructed, with significant outcomes, including RF, KNN, and MNB. The probabilities from MNB and Random Forest classifiers were combined to create a hybrid model that improved prediction accuracy. The hybrid model's greater accuracy and classification report, which highlight its potential for robust sentiment categorization, show that this strategy produced better results. The pipeline is effective and scalable for real-world applications because it combines preprocessing, feature extraction, and classification.