GeoPMF:距离敏感的旅游推荐模型

发布时间：2018-06-28 21:29

  本文选题：旅游推荐 + 推荐系统　； 参考：《计算机研究与发展》2017年02期

【摘要】：虽然目前旅游者可以利用Web搜索引擎来选择旅游景点,但往往难以获得较好符合自身需要的旅游规划.而旅游推荐系统是解决上述问题的有效方式.一个好的旅游推荐模型应具有个性化并能考虑用户时间和费用的限制.调研表明,用户在选择旅游景点时,目的地与用户常居地的距离常常是一个需要考虑的问题.因为旅行距离往往可以间接地反映了时间和费用的影响.于是,在贝叶斯模型和概率矩阵分解模型的基础上,提出一个旅行距离敏感的旅游推荐模型(geographical probabilistic matrix factorization,GeoPMF).主要思想是基于每个用户的旅游历史,推算出一个最偏好的旅游距离,并作为一种权重,添加到传统的基于概率矩阵分解的推荐模型中.在携程网站的旅游数据集上的实验表明,与基准方法相比,GeoPMF的RMSE(root mean square error)可以降低近10%;与传统概率矩阵分解模型(PMF)相比,通过考虑距离因子,RMSE平均降幅近3.5%.
[Abstract]:At present, tourists can use Web search engine to choose tourist attractions, but it is often difficult to obtain a better tourism planning that meets their needs. Tourism recommendation system is an effective way to solve the above problems. A good travel recommendation model should be personalized and take into account user time and expense constraints. The research shows that the distance between the destination and the user's place of residence is often a problem to be considered when users choose tourist attractions. Because travel distance can often indirectly reflect the impact of time and expenses. On the basis of Bayesian model and probability matrix decomposition model, a travel recommendation model (geographical probabilistic matrix factorization _ GeoPMF) is proposed. The main idea is to calculate a preferred travel distance based on each user's travel history and add it to the traditional recommendation model based on probability matrix decomposition as a weight. The experiment on the travel data set of Ctrip station shows that compared with the reference method, the RMSE (root mean square error) of GeoPMF can be reduced by nearly 10%, and the average decrease of RMSE by taking into account the distance factor is about 3.5% compared with the traditional probability matrix decomposition model (PMF).
【作者单位】： 山东大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(61272240,61672322) 山东省自然科学基金项目(ZR2012FM037) 微软国际合作基金项目(FY14-RESTHEME-25)~~
【分类号】：TP391.3
，

本文编号：2079410