FG-CF: Friends-aware graph collaborative filtering for POI recommendation

Cai, Zhuo; Yuan, Guan; Qiao, Shaojie; Song, Qian; Zhang, Yanmei; Bing, Rui

doi:10.1016/j.neucom.2022.02.070

Cited by 39 publications

(14 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to improve the accuracy of the predictions, data sparsity problem needs to be considered. Many previous works employ collaborative filtering (CF) technology and data sparsity problem can be alleviated [4,5] . The key idea of CF-based approach is to calculate similarities between different users, and estimate visiting preference of a certain user according to his similar users.…”

Section: Related Workmentioning

confidence: 99%

Gridding-based tensor decomposition algorithm for points of interest prediction in LBSNs

Du¹

2023

International Conference on Statistics, Data Science, and Computational Intelligence (CSDSCI 2022)

View full text Add to dashboard Cite

Location based social network (LBSN) provides large amounts of data which record the locations visited by users and corresponding latitude and longitude of these locations. Such datasets can be used to explore visiting preferences of users and predict the locations which are likely to be visited by a particular user in the future. Thus, the problem of prediction of users’ preference locations has become a research hotspot and attracts great attention from academia and practitioners. However, it is still a challenge to precisely predict which locations will be visited by users. The main reason is that the visiting decisions made by users will be affected by not only preferences but also geographical factors. In this paper, we investigate the influence of geographical factors, and propose a gridding-based tensor decomposition algorithm for users’ preference locations prediction. We divide the entire city into grids and fill these grids with visiting records of users. A tensor is constructed according these grids, and a tensor decomposition algorithm is employed to calculate the visiting probability of each grid for each user. Then, we calculate the popularity of locations in each grid. Finally, we construct a ranking list of all locations by considering both grids' visiting probabilities and corresponding popularity scores. We have implemented our algorithm and compared with existing approaches by using two public datasets, Foursquare and Gowalla. The experimental results show that our algorithm achieves higher precision and recall.

show abstract

Section: Related Workmentioning

confidence: 99%

Gridding-based tensor decomposition algorithm for points of interest prediction in LBSNs

Du¹

2023

International Conference on Statistics, Data Science, and Computational Intelligence (CSDSCI 2022)

View full text Add to dashboard Cite

show abstract

“…The importance of a friend [3] evaluates the influence of a friend while visiting the POI. Numerous studies [81][82][83][84][85][86] indicate that social relations are beneficial for the recommender systems, and the use of social factors to reinforce traditional recommendation systems has been investigated, both in memory-based methods [87,88] and in model-based techniques [89][90][91]. Attention to social influence and friend impression in POI selection has improved the recommendations of traditional recommender systems.…”

Section: • Social Influence and Importance Of Friends' Behaviormentioning

confidence: 99%

Toward Point-of-Interest Recommendation Systems: A Critical Review on Deep-Learning Approaches

2022

View full text Add to dashboard Cite

In recent years, location-based social networks (LBSNs) that allow members to share their location and provide related services, and point-of-interest (POIs) recommendations which suggest attractive places to visit, have become noteworthy and useful for users, research areas, industries, and advertising companies. The POI recommendation system combines different information sources and creates numerous research challenges and questions. New research in this field utilizes deep-learning techniques as a solution to the issues because it has the ability to represent the nonlinear relationship between users and items more effectively than other methods. Despite all the obvious improvements that have been made recently, this field still does not have an updated and integrated view of the types of methods, their limitations, features, and future prospects. This paper provides a systematic review focusing on recent research on this topic. First, this approach prepares an overall view of the types of recommendation methods, their challenges, and the various influencing factors that can improve model performance in POI recommendations, then it reviews the traditional machine-learning methods and deep-learning techniques employed in the POI recommendation and analyzes their strengths and weaknesses. The recently proposed models are categorized according to the method used, the dataset, and the evaluation metrics. It found that these articles give priority to accuracy in comparison with other dimensions of quality. Finally, this approach introduces the research trends and future orientations, and it realizes that POI recommender systems based on deep learning are a promising future work.

show abstract

“…It has been widely used in Web service QoS prediction. Its basic idea is that if users exhibit similar preferences with others in the past, they will have similar preferences in the future [8,[17][18][19][20]. Therefore, the essence of the collaborative filtering-based QoS prediction method is to find a set of similar users or services based on historical data, and then to infer the experience of the target user when invoking the target service by the performance of this user set or service set.…”

Section: Static Qos Predictionmentioning

confidence: 99%

Empirical Mode Decomposition with Multivariable Time Series Feature Learning for QoS Prediction

Zhao

Wang

et al. 2022

Preprint

View full text Add to dashboard Cite

Predicting the quality of service (QoS) for Web services has become an essential topic in services computing. Existing QoS prediction approaches typically treat each QoS record as a basic unit, ignoring that a QoS time series is a hybrid sequence with multiple frequency features. As a result, we propose AE-mLSTM, a hybrid QoS forecasting method that combines the empirical mode decomposition (EMD) model and the multivariable LSTM model, to capture the intrinsic properties of time-series QoS records. In addition, AE-mLSTM employs an attention mechanism for multi-task learning, which can learn the shared representation of different tasks and jointly predict the tasks of QoS and timing direction. Experiments conducted on two real-world datasets demonstrate that our approach outperforms several state-of-the-art QoS prediction methods. It can improve the average RMSE, MAPE, and MAE by 5.05%, 4.68%, and 5.74% for the Web service dataset and by 10.99%, 12.33%, and 10.93% for the IoT service dataset.

show abstract

FG-CF: Friends-aware graph collaborative filtering for POI recommendation

Cited by 39 publications

References 21 publications

Gridding-based tensor decomposition algorithm for points of interest prediction in LBSNs

Gridding-based tensor decomposition algorithm for points of interest prediction in LBSNs

Toward Point-of-Interest Recommendation Systems: A Critical Review on Deep-Learning Approaches

Empirical Mode Decomposition with Multivariable Time Series Feature Learning for QoS Prediction

Contact Info

Product

Resources

About