Comparative Analysis of LSTM and Random Forest Algorithms for Sentiment Classification in Movie Reviews

This study conducts sentiment analysis on 50,000 movie reviews, comparing the traditional Random Forest classifier with a Long Short-Term Memory (LSTM) network. The Random Forest achieved robust metrics across accuracy, precision, recall, and F1-score (0.85), while the LSTM excelled in capturing the complex language patterns inherent in the reviews through its sequence-based processing. This study has refined the data by removing irrelevant characters and employing tokenization and lemmatization to enhance model training. Despite its higher computational requirements, the LSTM showed potential in detailed sentiment analysis. Future directions include exploring other neural architectures, enlarging the dataset, and practical applications to improve model efficacy. The study highlights key insights into sentiment analysis methodologies and their implications for both academic and practical realms.