A Comprehensive Collaborating Filtering Approach using Extended Matrix Factorization and Autoencoder in Recommender System

Recommender system is an approach where users get suggestions based on their previous preferences. Nowadays, people are overwhelmed by the huge amount of information that is being present in any system. Sometimes, it is difficult for a user to find an appropriate item by searching the desired content. Recommender system assists users by providing suggestions of re-quired information or items based on the similar features among the users. Collaborative filtering is one of the most re-known process of recommender system where the recommendation is done by similar users or similar items. Matrix factorization is an approach which can be used to decompose a matrix into two or more matrix to generate features. Again, autoencoder is a deep learning based technique which is used to find hidden features of an object. In this paper, features are calculated using extended matrix factorization and autoencoder and then a new similarity metric has been introduced that can calculate the similarity efficiently between each pair of users. Then, an improvement of the prediction method is introduced to predict the rating accurately by using the proposed similarity measure. In the experimental section, it has been shown that our proposed method outperforms in terms of mean absolute error, precision, recall, f-measures, and average reciprocal hit rank.

@article{Hasan2019,
  title     = {A Comprehensive Collaborating Filtering Approach using Extended Matrix Factorization and Autoencoder in Recommender System},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {10},
  number    = {6},
  year      = {2019},
  publisher = {The Science and Information Organization},
  author    = {Mahamudul Hasan and Falguni Roy and Tasdikul Hasan and Lafifa Jamal},
  doi       = {10.14569/IJACSA.2019.0100666},
  url       = {https://doi.org/10.14569/IJACSA.2019.0100666}
}