Enabling “Untact” Culture via Online Product Recommendations: An Optimized Graph-CNN based Approach

Abstract

The COVID-19 pandemic is swiftly changing our behaviors toward online channels across the globe. Cultural patterns of working, thinking, shopping, and use of technology are changing accordingly. Customers are seeking convenience in online shopping. It is the peak time to assist the digital marketplace with right kind of tools and technologies that uses the strategy of click and collect. Session-based recommendation systems have the potential to be equally useful for both the customers and the service providers. These frameworks can foresee customer’s inclinations and interests, by investigating authentic information on their conduct and activities. Various methods exist and are pertinent in various situations. We propose a product recommendation system that uses a graph convolutional neural network (GCN)-based approach to recommend products to users by analyzing their previous interactions. Unlike other conventional techniques, GCN is not widely explored in recommendation systems. Therefore, we propose a variation of GCN that uses optimization strategy for better representation of graphs. Our model uses session-based data to generate patterns. The input patterns are encoded and passed to embedding layer. GCN uses the session graphs as input. The experiments on data show that the optimized GCN (OpGCN) was able to achieve higher prediction rate with around 93% accuracy as compared with simple GCN (around 88%).