Neural attention model for recommendation based on factorization machines
- PDF / 846,970 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 23 Downloads / 232 Views
Neural attention model for recommendation based on factorization machines Peng Wen 1 & Weihua Yuan 1 & Qianqian Qin 1 & Sheng Sang 1 & Zhijun Zhang 1 Accepted: 30 August 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract In recommendation systems, it is of vital importance to comprehensively consider various aspects of information to make accurate recommendations for users. When the low-order feature interactions between items are insufficient, it is necessary to mine information to learn higher-order feature interactions. In addition, to distinguish the different importance levels of feature interactions, larger weights should be assigned to features with larger contributions to predictions, and smaller weights to those with smaller contributions. Therefore, this paper proposes a neural attention model for recommendation (NAM), which deepens factorization machines (FMs) by adding an attention mechanism and fully connected layers. Through the attention mechanism, NAM can learn the different importance levels of low-order feature interactions. By adding fully connected layers on top of the attention component, NAM can model high-order feature interactions in a nonlinear way. Experiments on two real-world datasets demonstrate that NAM has excellent performance and is superior to FM and other state-of-the-art models. The results demonstrate the effectiveness of the proposed model and the potential of using neural networks for prediction under sparse data. Keywords Recommendation systems . Factorization machines . Attention weight . Feature interactions
1 Introduction Recommendation systems are essentially information filtering systems that learn users’ interests based on their information or historical behavior to recommend products for them [1]. It changes the communication between users and businesses and improves interaction and efficiency. Recommendation systems can be summarized as memory and generalization [2]. Memory is defined as the learning of common features of existing data and the mining of internal connections between the features. Generalization is defined as the exploration of feature combinations that have not yet occurred. At present, there is a large amount of available information and a high diversity of data forms [3]. Many enterprise-level recommendation systems use rich and varied data to mine different interactions between users and items from different perspectives [4]. Therefore, recommendation systems may not be effective if the results are based only on models of the original features. For example, the linear regression method
* Zhijun Zhang [email protected] 1
School of Computer Science and Technology, Shandong Jianzhu University, Jinan 250001, Shandong, China
assigns a weight to each feature, and the target is predicted as the weighted sum of all features. However, the results are unsatisfactory because the interactions between features are usually ignored. Therefore, it is important to effectively mine and interpret the interactions between features [5]
Data Loading...
