Enhancing Marketing Analytics in Online Retailing through Machine Learning Classification Techniques

Abstract

In the fiercely competitive retail industry, satisfying consumer expectations while optimizing company processes is more important than ever. Therefore, it is crucial to handle and channel data in a way that both seeks to delight consumers and generates healthy revenues if you want to survive and prosper. Data—or more specifically, Big data analytics is being utilized by large retailers at every stage of the process, participants in the global and Indian retail markets, including tracking new, popular items and predicting sales. The use of machine learning classification approaches for sentiment analysis in online shopping is examined in this research, utilizing a publicly available Amazon review dataset. The text-cleaning techniques processed the dataset before converting texts into numerical representations by implementing TF-IDF measures. The assessment concentrated on the three machine learning model's F1-score, accuracy, and precision-recall: Bidirectional Encoder Representations from Transformers (BERT), Support Vector Machine (SVM), and Gradient Boosting (GB). BERT ended up outperforming all other models by demonstrating 89% accuracy, which proves its extraordinary capability to detect customer sentiments. The research results show how transformer-based models work for improving sentiment analysis procedures in marketing analytics applications.