Building an Innovative E-Commerce Platform Powered by Session-Based Recommendations

Abstract

Businesses and consumers’ online shopping habits have been significantly altered by the rise of e-commerce in recent years. To assist users in finding products and streamlining the shopping experience, major platforms such as Amazon, eBay, and Alibaba employ recommendation algorithms. These algorithms are key to enhancing user satisfaction and driving sales. Although content-based approaches and collaborative filtering are popular traditional recommendation systems, they struggle to adapt to the rapidly shifting preferences of users. This is where Session-Based Recommendation Systems (SBRSs) come in. Without relying on long-term user data, SBRSs focus on predicting the next product a user will engage with based solely on their actions during the current session. This strategy is especially helpful on e-commerce platforms, where users often browse anonymously and make quick decisions. The purpose of this thesis is to use long short-term memory (LSTM) networks to enhance session-based recommendations in e-commerce. Due to their capacity to recognize patterns in user behavior over time, LSTM networks are ideal for this task. By predicting the next item a user will interact with based on session activity, these models significantly improve recommendation quality. To demonstrate their practical impact, an e-commerce platform called VIA was developed, showcasing how LSTM-based session-based recommendation models can deliver highly personalized shopping experiences