Experimental Hybrid Approach for Improving the Quality of Personalized Product Recommendation Systems with Deep Learning

Abstract

Listen

Translate article

Deep learning has recently made significant advancements in recommender systems. Various methods, including hybrid recommendation systems, content-based recommendations, and deep learning, have been employed to enhance recommendation accuracy. Over the past decade, the proliferation of big data has led to a rapid increase in the amount of available data on the internet. However, dealing with complex and large data sets makes it challenging for many users to quickly find the information they need. At this juncture, recommendation systems play a crucial role in addressing the issue of data overload. The development of recommendation algorithms has been particularly driven by the growth of the e-commerce industry. Traditional single recommendation algorithms face challenges such as data sparsity, long-tail items, and cold starts. Hybrid recommendation algorithms can effectively mitigate some of these limitations. In response to these challenges, this paper proposes an experimental hybrid approach to improve the quality of personalized product recommendations using deep learning, specifically an Improved Attention-Convolutional Neural Network (IA-CNN). This method aims to address the shortcomings of single collaborative models. The algorithm first employs a comprehensive approach, combining product- and user-based collaborative filtering strategies to generalize and categorize the output results. This methodology captures detailed and abstract nonlinear interactions between users and products through advanced deep learning techniques. Finally, experiments were designed to test the algorithm's effectiveness. The IA-CNN algorithm was evaluated against a benchmark algorithm using the Amazon product rating dataset. The results demonstrate that the proposed IA-CNN algorithm outperforms the benchmark in rating prediction on the test dataset