Sentimental analysis of customer product reviews to understand customer needs using machine learning

Abstract

In the digital age, user reviews have become pivotal in shaping businesses.The process of analyzing user reviews is challenging because of the large volumes of data and presence of spam. However, automated sentiment analysis can significantly enhance deeper understanding of customer needs and optimize marketing strategies. This study introduces a sentiment analysis framework leveraging machine learning, including deep learning models like CNNs and BERT, along with traditional approaches such as SVM and Logistic Regression. Using a customized dataset collected through web scraping, we address challenges in data preprocessing and feature extraction, ultimately providing insights to enhance marketing strategies and product development based on customer feedback. We scraped Amazon reviews in the computer accessories category using Beautiful Soup, focusing on data from 2021 onward. After collecting 36,792 reviews and removing redundancies, we finalized a dataset of 29,755 unique entries, capturing essential user details and ratings. Lastly, to interpret model performance, we incorporate the explainable AI method LIME. In our study, BERT demonstrates a balanced performance across all metrics, with accuracy (86.01%), recall (89%), F1 score (88%), and precision (87%), making it a superior choice for sentiment analysis.