Hybrid Recommendation Research Based on Ontology Semantics and Demographic Information
Download as PDF
DOI: 10.23977/CNCI2020011
Author(s)
Jianhong Jiang, Qianqian Li
Corresponding Author
Jianhong Jiang
ABSTRACT
To reduce the influence of data sparseness and cold start problem on recommendation, an improved hybrid recommendation algorithm based on ontology semantics and demographic information is introduced. The ontology structure is constructed by the hybrid recommendation model based on the item description. The item similarity results is strengthened by analyzing the semantic relevance of the ontology. Then combining with the recommendation method based on demographic information, and the two recommended algorithms are trained by BP neural network to achieve the optional weight for mixed recommendation. Finally, the Guilin tourism destination domain ontology is established and the web crawler is used to climb the data of user platform to train and test the recommendation algorithm. The mean absolute error (MAE) and the prediction coverage (COV) value are used as evaluation index to compared with other recommendation algorithms. The results illustrate that after adding demographic information to the recommendation based on the ontology similarity recommendation algorithm, the recommendation accuracy is improved and recommendation effect is more stable.
KEYWORDS
Ontology; semantic relevance; hybrid recommendation; demographic information; neural network