Enhanced Detection of Consumer Behavioral Shifts in E-Commerce Platforms with Transformer-Based Algorithms

This research aims to analyze changes in consumer behavior on e-commerce platforms using consumer interaction data such as view, add to cart, and purchase. Identifying changes in consumer behavior on e-commerce platforms is very important because it can provide deeper insight into consumer motivations and preferences. By better understanding how consumers interact with products, companies can design more targeted strategies to increase conversions, reduce cart abandonment, and improve the overall customer experience. The DistilBERT based prediction model is applied to detect and predict changing patterns of consumer behavior in the purchasing process. DistilBERT was chosen because of its more efficient capabilities compared to previous models which enable faster data processing and lower resource usage, which is very important for real-time applications on e-commerce platforms with big data. The data used includes consumer interactions during a certain period, with model evaluation using precision, recall, F1-score, and accuracy metrics. The results showed that despite an increase in the number of actions such as View and Add to Cart, conversion to Purchase was still hampered, indicating a cart abandonment problem. The model used managed to achieve 90% accuracy, with a precision value of 0.87, recall of 0.85, and F1-score of 0.86, showing excellent performance in predicting changes in consumer behavior. Based on the results of this analysis, companies can optimize marketing strategies by targeting consumers who have added products to their basket but have not yet made a purchase, as well as making price adjustments, discounts, and limited time offers. This research also emphasizes the importance of using real-time data to dynamically adjust marketing strategies and improve customer experience.