Image-Based Recommendation System using JPEG- Coefficient and RFs Approach

Abstract

Online shopping platforms are expanding at an unstoppable rate all over the world. These platforms mostly depend on search engines, which are still primarily based on the text-base and use keywords matching for finding similar products. However, customers want an interactive platform that would be easy, convenient and reliable for searching related products. In this paper, we have proposed a novel idea of searching for products on an online shopping platform using an image-based approach. In this, a user can provide, select, or click an image, and similar image-based products will be provided to the user. The proposed recommendation system is based on content-based image retrieval and is composed of two major phases; Phase 1 and Phase 2. In Phase 1, the proposed way would find the class or type of the product. In Phase 2, the recommendation system retrieves closely matched similar products. For Phase 1, the approach creates a model of products using Machine Learning (ML). Then the model is used to find the category of the test products. From the ML perspectives, we have used the Random Forests (RF) classifier, and for feature extraction, we have used the JPEG coefficients. The dataset worked upon here includes 20 categories of products. In Phase 1, the evaluation of the proposed model generates a 75% accurate model. For further enhancement of performance, the RF model has been integrated into the Deep Learning (DL) setup achieving 84% accurate predictions. Based on the customized evaluation approach for Phase 2, the proposed recommendation approach achieves 98% correct results, thus demonstrating its efficacy and accuracy for the product recommendation and searching in the daily life routine and practical applications.