基于知识图谱和用户长短期偏好的个性化景点推荐

Personalized attraction recommendation based on the knowledge graph and users’long-term and short-term preferences

基于序列化的推荐算法在多个领域取得了不错的效果,但仍存在一些问题,如没有考虑所有项与项之间的关系,推荐准确度会大大降低。因此提出一种基于知识图谱和用户长短期偏好(KG-ULSP)的个性化景点推荐方法。通过引入知识图谱,使用网络表示学习方法,学习景点的特征向量表示,使得具有相似结构和相似属性的景点在低维特征空间中的距离比较近,以此表示他们的高级语义特征。然后利用门控循环单元GRU对已学习到的景点特征向量进行序列化信息建模,进一步抽取景点的访问序列特征。另外,考虑到用户偏好可能随时间发生变化,KG-ULSP模型同时学习用户的长期偏好和短期偏好,最终预测并返回用户可能感兴趣的推荐列表。通过在真实旅游数据上的实验,验证了所提方法的有效性。

The session-based recommendation algorithm has achieved good results in many fields.However,several problems,such as not considering the relationship between all items,will reduce the recommendation accuracy considerably.Therefore,a personalized attraction recommendation method based on the knowledge graph and users’long-term and short-term preferences(KG-ULSP)is proposed.The knowledge graph is derived using the network representation learning method and the feature vector representation of the learning attractions.The attractions with similar structure and attribute are close to each other in the low-dimensional space and express high-level semantic features.In addition,the sequence information is modeled by the gated recurrent unit and the access sequence information is further extracted by feature extraction.Moreover,given that the users’preferences may change with time,the KG-ULSP model learns both long-term and short-term preferences of the user and predicts and returns the list of recommendations that users may be interested in.The validity of the proposed method is verified by experiments on real tourism data.