Hybrid Collaborative Recommendation via Dual-Autoencoder

Abstract

With the rapid increase of internet information, personalized recommendation systems are an effective way to alleviate the information overload problem, which has attracted extensive attention in recent years. The traditional collaborative filtering utilizes matrix factorization methods to learn hidden feature representations of users and/or items. With deep learning achieved good performance in representation learning, the autoencoder model is widely applied in recommendation systems for the advantages of fast convergence and no label requirement. However, the previous recommendation systems may take the reconstruction output of an autoencoder as the prediction of missing values directly, which may deteriorate their performance and cause unsatisfactory results of recommendation. In addition, the parameters of an autoencoder need to be pre-trained ahead, which greatly increases the time complexity. To address these problems, in this paper, we propose a Hybrid Collaborative Recommendation method via Dual-Autoencoder (HCRDa). More specifically, firstly, a novel dual-autoencoder is utilized to simultaneously learn the feature representations of users and items in our HCRDa, which obviously reduces time complexity. Secondly, embedding matrix factorization into the training process of the autoencoder further improves the quality of hidden features for users and items. Finally, additional attributes of users and items are utilized to alleviate the cold start problem and to make hybrid recommendations. Comprehensive experiments on several real-world data sets demonstrate the effectiveness of our proposed method in comparison with several state-of-the-art methods.

A Scalable Recommendation Engine for New Users and Items

“…Given the engineering challenges of updating item positions in a real-time environment, assessing the optimal level of updating (such as real time, minute, hour, or daily frequency) or how to batch recommendations is of interest. Second, hybrid recommendation methods have recently combined state-of-the-art representation learning techniques (e.g., autoencoders) with traditional heuristics (e.g., matrix factorization, logistic regression) (Jannach et al, 2020;Strub et al, 2016;Dong et al, 2020;Geng et al, 2022), suggesting that CFB-A could be improved by combining matrix factorization with autoencoders. Related and third, ours is a list recommender system.…”

“…However, model-based CF relies on rich historical records, a condition which is not satisfied in the contexts of new users and new items. Hence, our approach augments the model-based CF with user and item attributes to address the cold-start problem, similar to recent advances incorporating deep learning into CF (e.g., Dong et al, 2020;Geng et al, 2022).…”

Neural Collaborative Autoencoder for Recommendation With Co-Occurrence Embedding

“…Among them, pan et al [20] proposed TDAE model, which introduces social information and uses denoising autoencoder to learn compact representations, and a neural network is designed at the hidden layer to balance the importance of different representations. Dong et al [21] proposed HRDCa model, which utilizes additional attribute information to model the latent representation of users and items separately by dual autoencoder. Alfarhood et al [22] proposed CATA++ model, which focuses on scientific article recommendation problem, which employs two autoencoders and attention mechanism to capture significant contextual information.…”

An improved hybrid semi-stacked autoencoder for item-features of recommendation system (iHSARS)

“…With the explosion of information on the Internet in recent years, the problem of information overload arises [1]. Therefore, it is increasingly difficult to efficiently filter relevant information from all the data available on the internet, which is a potential challenge for many Internet users [2].…”