DeepAPF: Deep Attentive Probabilistic Factorization for Multi-site Video Recommendation

Abstract

Existing web video systems recommend videos according to users' viewing history from its own website. However, since many users watch videos in multiple websites, this approach fails to capture these users' interests across sites. In this paper, we investigate the user viewing behavior in multiple sites based on a large scale real dataset. We find that user interests are comprised of cross-site consistent part and site-specific part with different degrees of the importance. Existing linear matrix factorization recommendation model has limitation in modeling such complicated interactions. Thus, we propose a model of Deep Attentive Probabilistic Factorization (DeepAPF) to exploit deep learning method to approximate such complex user-video interaction. DeepAPF captures both cross-site common interests and site-specific interests with non-uniform importance weights learned by the attentional network. Extensive experiments show that our proposed model outperforms by 17.62%, 7.9% and 8.1% with the comparison of three state-of-the-art baselines. Our study provides insight to integrate user viewing records from multiple sites via the trusted third party, which gains mutual benefits in video recommendation.

Attraction Recommendation Based on Tourism Context Modeling and Multi‐neural Collaborative Filtering Algorithm

“…However, the model has a high square-level complexity for the number of non-zero features. Yan et al [22] proposed the deep attentive probabilistic factorization. The model learns nonuniform importance weights through attention networks to accurately capture user interest across and site-specific sites.…”

A Survey on Cross-domain Recommendation: Taxonomies, Methods, and Future Directions

“…Some researchers improved previous works by considering users' domain-shared representations and domainspecific representations. Yan et al [92] proposed a model of Deep Attentive Probabilistic Factorization (DeepAPF) in which the user embeddings were initialized into three parts: 𝑃 𝑆 𝑈 , 𝑃 𝐴 𝑈 , and 𝑃 𝐵 𝑈 . 𝑃 𝑆 𝑈 was the domain-shared representations capturing cross-domain commonality of user interests and 𝑃 𝐴 𝑈 , 𝑃 𝐵 𝑈 were the domain-specific representations capturing site-peculiarity of user interests.…”

“…Traditional machine learning methods, such as matrix factorization [85,91], factorization machines [56,62], co-clustering [50,71,90], and latent semantic analysis [63,87] have been widely applied. In recent years, with the emergence and development of deep learning technologies, many approaches based on deep learning have been proposed [27,33,38,39,65,92,104], which greatly improves the accuracy and performance of cross-domain recommendation. To answer the question of how to transfer, a straightforward idea is to utilize overlapping entities, either users or items, to directly establish relationships between domains [23,40,77,85,92,109,111].…”

“…In recent years, with the emergence and development of deep learning technologies, many approaches based on deep learning have been proposed [27,33,38,39,65,92,104], which greatly improves the accuracy and performance of cross-domain recommendation. To answer the question of how to transfer, a straightforward idea is to utilize overlapping entities, either users or items, to directly establish relationships between domains [23,40,77,85,92,109,111]. When there are no overlapping entities, some efforts are also made to establish linkages by extracting cluster-level patterns [50,71,84,90] or resorting to other auxiliary information (e.g., users' generated tags, reviews, user profiles, and item content) [30,83,96,98].…”

Explicitly Modeling Relationships between Domain-Specific and Domain-Invariant Interests for Cross-Domain Recommendation

“…To overcome this problem, researchers proposed that both users' domain-invariant interests and domain-specific interests should be modeled simultaneously. Specifically, Yan et al proposed the DeepAPF [13] method that leverages an attentional network to learn non-uniform importance weights of users' domain-invariant and domain-specific representations. Xu et al proposed the ReCDR [14] method which constructs both single-domain and cross-domain graphs to generate users' two kinds of representations.…”