Content Based Apparel Recommendation Engine

Abstract

With the rapid rising of living standard,  people gradually developed higher shopping enthusiasm  and increasing demand for garment. Nowadays, an  increasing number of people pursue fashion. However,  facing too many types of garment, consumers need to try  them on repeatedly, which is somewhat time and energy consuming. Besides, it is difficult for merchants to master  the real- time demand of consumers. Proposed  recommendation engine utilizes tops_fashion dataset which  consists nearly 183k products acquired through Amazon  API interface and processes this dataset using multiple  techniques like Bag_of_Words (BoW), tf, idf, Word2Vec  model,VGG16 (CNN) etc to recommend similar items to a  given query item. Product Advertising API acts as a  gateway to Amazon's databases so that we can take  advantage of Amazon's sophisticated e-commerce data and  functionality. In this project, we are using Python language  for coding. The proposed system can recommend product  based on various features of the product such as title, color,  brand, price, image but we have only used title and image  features of the products. Collaborative Filtering technique  suffers from cold start problem- a situation where a  recommender does not have adequate information about a  user or an item in order to make relevant predictions. This is  one of the major problems that reduce the performance of  recommendation system.