Name
Abstract | ||
|---|---|---|
Human mobility prediction is of great significance for analyzing the check-in data generated by location-based applications. Compared with classical prediction methods, recently published ones based on neural networks have made significant improvements, but there still exist problems. First, several valuable characteristics in human mobility, such as the geographic relevance, community, and diversity of user movements, are not fully exploited. Second, the sparsity and imbalance of the check-in data also greatly restrict the prediction performance. To alleviate them, this manuscript proposes a new human mobility prediction method called the group-based multi-features move (GMFMove). This method constructs a prediction model based on recurrent neural network and attention mechanism. Three important factors that influence user movements, i.e., the sequence of location, the category of location, and the geographic relevance of human mobility, are taken into consideration in the model to better capture the mobility preference. Furthermore, GMFMove uses a deep-learning-based matrix factorization to integrate prior information including implicit feedbacks and social relationships for grouping users. Then, for each user group, we use a separate multi-features move (MFMove) model to train it and get the subresult. Finally, all of them are integrated in terms of the weights to obtain the final prediction result. We conduct extensive experiments on four real check-in datasets, and the experimental results show that GMFMove method significantly outperforms other methods. |
Year | DOI | Venue |
|---|---|---|
2022 | 10.1007/s00521-022-06971-6 | Neural Computing and Applications |
Keywords | DocType | Volume |
Human mobility prediction, Geographic relevance, Recurrent neural network, Group based, Deep learning, Matrix factorization | Journal | 34 |
Issue | ISSN | Citations |
12 | 0941-0643 | 0 |
PageRank | References | Authors |
0.34 | 6 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Shengren Ke | 1 | 0 | 0.34 |
| Meiyi Xie | 2 | 7 | 4.86 |
| Hong Zhu | 3 | 7 | 5.52 |
| Zhongsheng Cao | 4 | 2 | 1.37 |