Preliminary data-based matrix factorization approach for recommendation

Yuan, Xiaofeng; Han, Lixin; Qian, Subin; Zhu, Licai; Zhu, Jun; Yan, Hong

doi:10.1016/j.ipm.2020.102384

Cited by 21 publications

(7 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To improve the performance of these methods, some deep neural network-based recommendation methods have been proposed. For example, Neural MF [16], DeepFM [18] and PDMF [63] leverage neural networks [3,6] to infer users' preferences from historic user-item interactions based on the collaborative ltering hypothesis [52]. These methods have achieved outperformance and have been deployed in many real-world scenarios [11,13,71].…”

Section: Traditional Recommendation Systemsmentioning

confidence: 99%

Enhancing Conversational Recommendation Systems with Representation Fusion

et al. 2023

View full text Add to dashboard Cite

Conversational Recommendation Systems (CRSs) aim to improve recommendation performance by utilizing information from a conversation session. A CRS first constructs questions and then asks users for their feedback in each conversation session in order to refine better recommendation lists to users. The key design of CRS is to construct proper questions and obtain users’ feedback in response to these questions so as to effectively capture user preferences. Many CRS works have been proposed; however, they suffer from defects when constructing questions for users to answer: (1) employing a dialogue policy agent for constructing questions is one of the most common choices in CRS, but it needs to be trained with huge corpus; (2) it is not appropriate that constructing questions from a single policy (e.g., A CRS only selects attributes that the user has interacted with) for all users with different preferences. To address these defects, we propose a novel CRS model, namely, a R epresentation F usion-based C onversational R ecommendation model (RFCR), where the whole conversation session is divided into two subsessions (i.e., Local Question Search subsession and Global Question Search subsession) and two different question search methods are proposed to construct questions in the corresponding subsessions without employing policy agents. In particular, in the Local Question Search subsession we adopt a novel graph mining method to find questions, where the paths in the graph between users and attributes can eliminate irrelevant attributes; in the Global Question Search subsession we propose to initialize user preference on items with the user and all item historical rating records and construct questions based on user’s preference. Then, we update the embeddings independently over the two subsessions according to user’s feedback and fuse the final embeddings from the two subsessions for the recommendation. Experiments on three real-world recommendation datasets demonstrate that our proposed method outperforms five state-of-the-art baselines.

show abstract

Section: Traditional Recommendation Systemsmentioning

confidence: 99%

Enhancing Conversational Recommendation Systems with Representation Fusion

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The efficiency of such rankingbased systems crucially depends on how they deal with sparse and missing entries in the user-item rating matrix [2]. Data imputation-based methods are proposed to overcome rating sparsity without employing auxiliary information [5][6][7]. Yuan et al [12] proposed the imputation-based SVD (ISVD) approach to generate and then include imputed ratings into the SVD model by inferring reliable neighbors for users and items.…”

Section: Related Workmentioning

confidence: 99%

“…Yuan et al [12] proposed the imputation-based SVD (ISVD) approach to generate and then include imputed ratings into the SVD model by inferring reliable neighbors for users and items. The PDMF imputation method [5] produces preliminary data to constrain the learning in MF. Although only selected neighbors' ratings are employed for imputation in ISVD and the correlations between the learned original, preliminary, and concatenated preferences are examined in PDMF, such neighborhood-based techniques do not estimate uncertainty in rating matrix's empty cells to impute missing values only to relevant positions [1].…”

Section: Related Workmentioning

confidence: 99%

“…However, in real-world applications, the user-item interaction matrices are sparse, with only a few observed rating values given by users on items [2,3]; hence using an MF technique to linearly model interaction data directly on a sparse rating matrix cannot deal with the non-linear and complex intrinsic structure of user and item latent features [2]. Imputation approaches are proposed to tackle sparsity issues by ensuring pre-processing for missing entries in the user-item matrix before generating ranked items for the target user [5][6][7]. Nevertheless, many data imputation-based models assume that unobserved rating information is missing at random (MAR) [8], considering that the probability of observing a rating does not rely on its value.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

“Guess Why I didn't rate it”: A New Preference-based Model for Enhanced TopK Recommendation

2023

IJIES

View full text Add to dashboard Cite

In existing matrix factorization (MF)-based recommender systems, the user-item interaction matrix is factorized linearly into two low-ranked feature matrices to generate predictions or provide the users with personalized rankings of items. However, when MF methods are directly applied to sparse rating matrices, they cannot cope with the inherent structure of real-world user and item latent features. The efficiency of MF-based systems crucially depends on their capability to address sparsity issues. To this end, we propose, in this paper, a novel preference-based data imputation approach for effective MF-based Top-K recommendation. We apply MF on an imputed and denser rating matrix with only interesting items to users. We obtain these items by inferring the prior preferences of users, considering some biases that may impact their choices for items they interact with, and leveraging a powerful latent and non-linear feature extraction using a deep generative model. Experimental results on two real-world sparse datasets reveal that the proposed model significantly enhances the performance of Top-K recommendations and outperforms baselines with an average improvement margin of 6.45% and 3.91% in hit rate (HR) and normalized discounted cumulative gain (NDCG) evaluation metrics, respectively, averaging on all employed datasets.

show abstract

“…However, deriving user preference is usually a difficult task, since even users themselves could not accurately articulate their real interests. In this circumstance, matrix factorization (MF) techniques (Gemulla et al, 2011;Huang et al, 2018;Immer et al, 2020;Kawale et al, 2015;Koren et al, 2009;Lee and Seung, 2000;Luo et al, 2014;Park et al, 2017;Rendle et al, 2020;Salakhutdinov and Mnih, 2007;Sorkunlu et al, 2018;Tran et al, 2018;Trigeorgis et al, 2017;Wang and Ma, 2020;Wu et al, 2020Wu et al, , 2021aYuan et al, 2021;Zeng et al, 2015;Zhang and Chow, 2016;Zhang et al, 2021) have been widely applied to recommendation systems to discover latent characteristics from explicit user feedback which implicitly describes the features of users and items. MF decomposes the user rating matrix on items into two separate matrices in a shared latent space; one matrix depicts the user vectors and the other depicts the item vectors.…”

Section: Introductionmentioning

confidence: 99%

A novel virtual-communicated evolution learning recommendation

Chen,

Chen

2023

IMDS

View full text Add to dashboard Cite

PurposeIn this “Info-plosion” era, recommendation systems (or recommenders) play a significant role in finding interesting items in the surge of online digital activity and e-commerce. The purpose of this paper is to model users' preference evolution to recommend potential items which users may be interested in.Design/methodology/approachA novel recommendation system, namely evolution-learning recommendation (ELR), is developed to precisely predict user interest for making recommendations. Differing from prior related methods, the authors integrate the matrix factorization (MF) and recurrent neural network (RNN) to effectively describe the variation of user preferences over time.FindingsA novel cumulative factorization technique is proposed to efficiently decompose a rating matrix for discovering latent user preferences. Compared to traditional MF-based methods, the cumulative MF could reduce the utilization of computation resources. Furthermore, the authors depict the significance of long- and short-term effects in the memory cell of RNN for evolution patterns. With the context awareness, a learning model, V-LSTM, is developed to dynamically capture the evolution pattern of user interests. By using a well-trained learning model, the authors predict future user preferences and recommend related items.Originality/valueBased on the relations among users and items for recommendation, the authors introduce a novel concept, virtual communication, to effectively learn and estimate the correlation among users and items. By incorporating the discovered latent features of users and items in an evolved manner, the proposed ELR model could promote “right” things to “right” users at the “right” time. In addition, several extensive experiments are performed on real datasets and are discussed. Empirical results show that ELR significantly outperforms the prior recommendation models. The proposed ELR exhibits great generalization and robustness in real datasets, including e-commerce, industrial retail and streaming service, with all discussed metrics.

show abstract

Preliminary data-based matrix factorization approach for recommendation

Cited by 21 publications

References 29 publications

Enhancing Conversational Recommendation Systems with Representation Fusion

Enhancing Conversational Recommendation Systems with Representation Fusion

“Guess Why I didn't rate it”: A New Preference-based Model for Enhanced TopK Recommendation

A novel virtual-communicated evolution learning recommendation

Contact Info

Product

Resources

About