Abstract

Sentiment analysis is an automated technique used to determine customers' opinions regarding products or services. Despite existing advancements, current models for analysing movie sentiments require further refinement to achieve more accurate and reliable sentiment classification. The study therefore, presents an improved movie review sentiment analysis model, designed based on Convolution Neural Network (CNN) architecture and optimized using Term Frequency-Inverse Document Frequency (tf-idf) algorithm. The performance evaluation of the optimized sentiment analysis model was carried out through comparative analysis with a conventional CNN model, using performance metrics, including accuracy, precision, recall, and F1-score to assess model performance. The simulation was driven by a robust movie dataset sourced from Kaggle. This result suggests that Opt-CNN is better in term of reliability for sentiment classification, thereby offering enhanced reliability.