Hybrid E-Recommendation System for Multi-Shop Environment

Abstract

In the Kurdistan Regional Government, most computer shops and markets conduct their marketing offline and do not have electronic systems. Nevertheless, customers live in a digital age; they often face challenges in finding products among these markets and shops. The most common question that customers ask is which shop they should purchase from. Therefore, data from five laptop stores and ratings for markets were collected to build an integrated recommender system to help customers find products and select the best store. Our proposed system is a hybrid e-recommendation system that combines machine learning techniques to provide personalized shop and product recommendations. Methods include data collection from multiple laptop shops and dataset preparation. The system uses techniques such as hybrid/blended methods using singular value decomposition and K-nearest neighbors for collaborative filtering (CF) to recommend shops and products based on customer ratings, alongside term frequency-inverse document frequency vectorization and cosine similarity for content-based filtering. The CF’s performance was evaluated using metrics like RMSE = 0.14 and MAE = 0.11, which demonstrated positive results for product and market recommendation. Overall, this study offers solutions through HE-RS to address key challenges such as market fragmentation, cold-start problems, and data scarcity.