A Comprehensive Collaborative Filtering Approach using Autoencoder in Recommender System
Mahamudul Hasan, Md. Tasdikul Hasan, Md. Selim Reza, Md. Nirab Akonda, M. Saddam Hossain Khan, Md. Mohsin Uddin · 2019
Recommender System is such kind of method where a user can get a recommendation by analyzing the user's previous preferences or behavior. It is an approach which helps a user to find items and contents by predicting their rating and showing them the recommended products. Items in the user based model are recommended to a user based on his/her similar user's preferences. In this paper, a strategy has been proposed in which calculation of the similarity between users have been done by using Autoencoder (AE) feature on the movielens data set. By using the autoencoder user-features have been calculated. Two users might like the same type of products and might give the same ratings to common products, so their features should be identical. By taking this relevant information into account, the similarity between users has been estimated and a collaborative filtering system based on user has been proposed. A visualization of the results of the recommendation system after using the evaluation methods have also been provided.