融合神经网络与矩阵分解的旅游景点推荐模型

Tourism Attraction Recommendation Algorithm Based on Deep Neural Network Matrix Factorization

针对旅游业不断发展、旅游有关信息过载、用户难以选择合适旅游景点等现实问题,研发了融合神经网络与矩阵分解的Tr-DNNMF旅游景点推荐模型。该模型将采用神经协同过滤模型的框架,使用广义矩阵分解模型和多层神经网络模型作为预训练模型单独训练,再对这两个模型进行融合,预测用户对景点的评分并进行景点推荐。本模型将矩阵分解的线性和深度神经网络的非线性相结合,对用户-景点的潜在特征进行建模,使模型既具备良好的扩展性,又有强大的拟合能力。利用国内旅游景点用户交互数据进行实验,实验结果表明:融合模型比单独模型的NDCG@10提高了7%,HR@10提高了0.5%。

In view of the continuous development of tourism,the overload of tourism related information,and the difficulty for users to choose suitable tourist attractions,this paper develops a TR NeuMF tourist attractions recommendation model combining neural network and matrix decomposition.The model will adopt the framework of the neural collaborative filtering model,use the generalized matrix factorization model and the multilayer neural network model as the pre-training model to train separately,and then integrate the two models to predict the user's rating of the scenic spot and recommend the scenic spot.This model combines the linearity of the matrix decomposition and the nonlinearity of the deep neural network to model the potential characteristics of the user and the scenic spot,so that the model has both good scalability and strong fitting ability.This paper uses the user interaction data of domestic tourist attractions to conduct experiments.The experimental results show that the fusion model in this paper is 7%higher than the single model NDCG@10,and HR@10 increases 0.5%.