基于贝叶斯网络认知反馈的协同过滤推荐

发布时间：2018-04-13 07:44

本文选题：协同过滤 + 推荐模型　；参考：《控制工程》2017年07期

【摘要】：协同过滤模型和算法是电子商务平台推荐系统的主要方法,存在数据稀疏性和推荐精准性问题;另外,推荐系统也面临推荐服务的满意度评价问题。为此,提出基于贝叶斯网络和用户认知的动态反馈模型,首先利用贝叶斯网络模型计算用户-商品的匹配概率,对商品属性评分数据降维,减少数据稀疏性;其次构建商品推荐反馈模型,定义用户认知函数和商品属性权重,确定商品多属性因素影响条件下的权重度量方法,同时保留用户的认知评价信息,定义用户满意度评价指标商品推荐率,改进用户-商品评价相似度计算方法;最后给出推荐算法的流程。通过应用实例和两类用户评价数据集进行多种算法的实验验证比较,结果表明:模型与其他协同过滤算法相比,具有较高的用户推荐满意度和推荐预测精度。
[Abstract]:Collaborative filtering model and algorithm are the main methods of E-commerce platform recommendation system, which have the problems of data sparsity and recommendation accuracy. In addition, recommendation system also faces the problem of satisfaction evaluation of recommendation service.For this reason, a dynamic feedback model based on Bayesian network and user cognition is proposed. Firstly, the Bayesian network model is used to calculate the matching probability of user and commodity, to reduce the dimension and reduce the data sparsity.Secondly, we construct the commodity recommendation feedback model, define the user cognitive function and commodity attribute weight, determine the weight measurement method under the condition of commodity multi-attribute factors, and retain the user's cognitive evaluation information at the same time.This paper defines the index of customer satisfaction evaluation commodity recommendation rate and improves the similarity calculation method of user-commodity evaluation. Finally the flow chart of recommendation algorithm is given.The experimental results show that the model has higher user recommendation satisfaction and recommendation prediction accuracy than other collaborative filtering algorithms.
【作者单位】：东北大学计算机科学与工程学院;东北大学工商管理学院;
【基金】：教育部留学回国人员科研启动基金资助项目(49-1) 国家级大学生创新创业训练计划资助项目(201510145011)
【分类号】：TP391.3

【相似文献】