IMAGE-BASED RECOMMENDATION SYSTEM.

Abstract

Online shopping has transformed consumer behavior, making online purchases more accessible. This shift has increased the need for effective and precise product recommendation systems. Traditional recommendation methods like Collaborative and Demographic Recommendation System face challenges, particularly the cold-start problem of suggesting products to new users with limited interaction histories. Content-Based Recommendation System, using keyword matching, often leads to inconsistent recommendations due to the varied descriptions of products. To address these challenges, this study introduces an Image-based recommendation system which is then combined with collaborative recommendation system to create a hybrid recommendation system. This method leverages the visual features of products in the dataset, such as color and shape, which are significant factors in consumer preferences, yet are often overlooked in textual descriptions. In addition, the dataset has been chosen and collected on the Internet and then stored at Kaggle [1].