Active Learning Empowered Sentiment Analysis: An Approach for Optimizing Smartphone Customer’s Review Sentiment Classification
Sidra Abbas, Wadii Boulila, Maha Driss, Gabriel Avelino Sampedro, Mideth Abisado, Ahmad Almadhor
Abstract
Product reviews are critical in informing customers about products and services and are crucial to customer inclinations and business success. Active learning enables Artificial Intelligence (AI) algorithms to learn more efficiently by actively selecting the most informative data samples for training. This study proposes a state-of-the-art approach that utilizes active learning (AL) based machine learning algorithms for Flipkart’s smartphone customers’ reviews sentiments analysis. The proposed study adopts rigorous data cleaning, data balancing, and Term Frequency-Inverse Document Frequency (TF-IDF) feature extraction techniques, followed by five AL-based machine learning algorithms, including Random Forest (AL-RF), Decision Tree (AL-DT), K-Nearest Neighbor (AL-KNN), Logistic Regression (AL-LR), and Gradient Boost (AL-GB), to classify sentiment analysis based on product reviews. The experiment spanned over three iterations with a chunk size of 200, and an active learning strategy was applied to the process. The results demonstrate the effectiveness of active learning in improving the accuracy of sentiment analysis on unbalanced datasets. Logistic Regression has demonstrated an accuracy rating of 89%. Collectively, the study showcases the potential of active learning techniques for sentiment analysis in product reviews and serves as a valuable tool for guiding customers in making informed decisions regarding their choice of products.