Research on Tourism Resources Management Method Based on Deep Learning and Knowledge Graph

被引:0
|
作者
Yang, Ling [1 ]
Huang, Xin [1 ]
机构
[1] Anhui Vocat Coll Grain Engn, Dept Bussiness Adm, 2 Xuelin Rd,High Educ Base, Hefei 230011, Anhui, Peoples R China
来源
2022 6TH INTERNATIONAL SYMPOSIUM ON COMPUTER SCIENCE AND INTELLIGENT CONTROL, ISCSIC | 2022年
关键词
GRU; knowledge graph; feature vector; tourism resources;
D O I
10.1109/ISCSIC57216.2022.00036
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In order to improve the efficiency of tourism resources management and the accuracy of tourism resources recommendation, a personalized recommendation method for tourism resources based on deep learning and knowledge graph was proposed. Firstly, the feature vectors of attractions were represented based on the Node2Vec method in knowledge graph feature learning. Then, the potential feature vectors of attractions were obtained by using the feature vectors of attractions as the input of the GRU neural network. Considering the user's long and short term preferences, the attention mechanism was introduced for modeling, and the probability of each attraction in the next visit was predicted, and fmally a reconunendation list was generated. The experimental results showed that the HR@10 index value of the proposed method is 41.1%, and the MRR@10 index value is 16.5%, which is higher than that of other reconunendation methods. At the same time, in the influence of different dimensions on the reconunendation results, with the increase of data dimensions, the hit rate of recommendations also increases. When the dimension is 150, the recommendation effect is the best. Among the effects of different reconunendation list lengths on reconunendation results, the proposed reconunendation method is higher than the other recommendation methods. Therefore, the proposed reconunendation method can well recommend tourism resources to users, and the reconunendation model shows good recommendation performance for multi-dimensional complex data.
引用
收藏
页码:127 / 131
页数:5
相关论文
empty
未找到相关数据