Towards latent context-aware recommendation systems

Unger, Moshe; Bar, Ariel; Shapira, Bracha; Rokach, Lior

doi:10.1016/j.knosys.2016.04.020

Cited by 110 publications

(46 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…General approaches such as the context neighbor recommender [12] and the general factorization framework [14] are also designed for directly modeling contexts. In contrast to using original contexts, some researchers recommend choosing important contexts to improve model performance [26], whereas other researchers improve the accuracy by extracting hidden patterns from the original contexts [27]. In addition to those single models, ensemble learning [28]- [31] can be useful in CAR.…”

Section: A New User Problemmentioning

confidence: 99%

Factorization Meets Neural Networks: A Scalable and Efficient Recommender for Solving the New User Problem

et al. 2020

View full text Add to dashboard Cite

Predicting new user behavior has always been a challenging issue in intelligent recommender systems. This challenge is mainly due to the extreme asymmetry of information between new users and old users. Existing factorization models can efficiently process and map asymmetric information, but they are not good at mining deep relationships between contexts when compressing high-dimensional data. In contrast, neural network methods can deeply exploit the relationship between contexts; however, their training cost is much larger than factorization approaches. Therefore, this paper proposes a scalable and efficient recommender to solve the new user problem by filling the gap between factorization models and neural networks. The scalable part is a neural network that can jointly encode, compress, and fuse various types of contexts. The efficient part is a factorization model based on a correlation constraint mechanism and a projection strategy, which enables an asymmetric mapping of information between old and new users. The entire recommender fuses the two parts so that the factorization model and the neural network can complement each other. The experimental results show that our approach can achieve a good balance between performance and training efficiency compared to state-of-the-art methods. INDEX TERMS Recommender systems, new user problems, user behavior, context-aware recommendation, efficient training.

show abstract

Section: A New User Problemmentioning

confidence: 99%

Factorization Meets Neural Networks: A Scalable and Efficient Recommender for Solving the New User Problem

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Choi et al [8] predicated a user's ratings on items according to a given user's purchase behaviors. Moshe et al [9] suggested an approach that was centered on representing environmental features as low dimensional unsupervised latent contexts. The latent contexts were automatically learned for each user utilizing unsupervised deep learning techniques and principal component analysis.…”

Section: Related Workmentioning

confidence: 99%

“…Tsim kj (9) The prediction of user u 's rating on tag k is determined by Formula (9) after tag neighbor selection, where NT j is top K neighbors of tag j . RT u = tag j r uj !…”

Section: Tag-driven Recommendation With Collaborative Item Modelingmentioning

confidence: 99%

Tag-Driven Online Novel Recommendation with Collaborative Item Modeling

Lin

Wang

2018

Information

View full text Add to dashboard Cite

Online novel recommendation recommends attractive novels according to the preferences and characteristics of users or novels and is increasingly touted as an indispensable service of many online stores and websites. The interests of the majority of users remain stable over a certain period. However, there are broad categories in the initial recommendation list achieved by collaborative filtering (CF). That is to say, it is very possible that there are many inappropriately recommended novels. Meanwhile, most algorithms assume that users can provide an explicit preference. However, this assumption does not always hold, especially in online novel reading. To solve these issues, a tag-driven algorithm with collaborative item modeling (TDCIM) is proposed for online novel recommendation. Online novel reading is different from traditional book marketing and lacks preference rating. In addition, collaborative filtering frequently suffers from the Matthew effect, leading to ignored personalized recommendations and serious long tail problems. Therefore, item-based CF is improved by latent preference rating with a punishment mechanism based on novel popularity. Consequently, a tag-driven algorithm is constructed by means of collaborative item modeling and tag extension. Experimental results show that online novel recommendation is improved greatly by a tag-driven algorithm with collaborative item modeling.

show abstract

“…Unger et al [7] suggested using high dimensional sensors, representing users' context for a CARS (Context-Aware Recommender System), in order to improve the quality of information gathered from many information sources that differ in the quality of information. Their proposed method can reduce the dimensionality by extracting latent contexts from data collected by mobile device sensors.…”

Section: Related Workmentioning

confidence: 99%

Toward Fashion-Brand Recommendation Systems Using Deep-Learning: Preliminary Analysis

Wakita¹,

Oku²,

Kawagoe³

2016

IJKE

View full text Add to dashboard Cite

Abstract-Recently, the number of Electronic Commerce users has been rapidly increasing with the spread of the Internet. However, users cannot easily find their preferred clothes items among the enormous number on the Internet. As a method for solving this problem, we propose a fashion-brand recommendation system using a deep learning method. This system increases the likelihood that a user will find his/her favorite clothes items. The user must first determine his/her favorite fashion-brands. In this paper, we evaluate the effectiveness of using a deep learning method in a fashion-brand recommendation system. The preliminary analysis shows that the fashion-brand recommendation method using deep learning can dramatically improve the recommendation accuracy as compared with other machine learning methods.

show abstract

Towards latent context-aware recommendation systems

Cited by 110 publications

References 15 publications

Factorization Meets Neural Networks: A Scalable and Efficient Recommender for Solving the New User Problem

Factorization Meets Neural Networks: A Scalable and Efficient Recommender for Solving the New User Problem

Tag-Driven Online Novel Recommendation with Collaborative Item Modeling

Toward Fashion-Brand Recommendation Systems Using Deep-Learning: Preliminary Analysis

Contact Info

Product

Resources

About