A ResNet-101 Based Recommendation System for E-Commerce

In today's connected world, where users have millions of choices on online platforms, recommendation systems play an important role in personalizing experiences. Traditionally, recommendation systems are created by utilizing collaborative or content-based filtering; however, both do not utilize visual aspects of items, which playa very important role in helping users to make decisions. With advances in deep learning techniques, image processing and extracting valuable information from images using convolutional neural networks are becoming easier and efficient. Therefore, images of items can provide very accurate and important auxiliary information for recommendation systems to utilize and make more accurate recommendations. This paper proposes a ResNet-lOl based recommendation system which considers the visual aspects as well as the metadata of items to increase the rating prediction performance of already existing recommendation algorithms. This paper also proposes a way through which recommendations could be generated by utilizing predicted ratings.