【作者】 岳文君；

【导师】 高强；

【作者基本信息】 北京邮电大学 ， 计算机科学与技术， 2013， 硕士

【摘要】 推荐系统是为缓解“信息爆炸”和“信息过载”现象这个问题而产生的一种信息服务技术,它根据用户历史行为信息来构建用户兴趣模型并通过该模型向用户推荐其可能感兴趣的信息。推荐系统能够在帮助用户的同时提高企业的利润,具有良好发展和应用前景,目前已经成为国内外学者的一个重要研究方向。为了能够更好的提供推荐,本文对融合位置上下文信息的协同过滤推荐算法做了探索性的研究；并将其应用于孕前检查评估建议推荐系统中。本文选题自“国家免费孕前优生健康检查信息服务平台”,在服务人员为服务对象提供评估建议的时候提供智能的评估建议推荐,辅助服务人员进行高质量的工作。本论文的主要研究内容如下：(1)本论文对智能推荐系统中的用户建模技术与推荐算法等关键技术进行了研究,对不同种类的推荐算法进行了总结与比较。(2)针对传统协同过滤的推荐系统在进行推荐是没有考虑用户地点上下文信息这一情况,本文提出并设计了一个基于地点上下文信息的协同过滤推荐算法。首先根据系统用户的位置信息进行距离衰减度的计算,再基于用户之间的兴趣行为相似性构建用户偏好关系网络；通过两者的结合,得到用户之间的相似度；最后通过为用户进行推荐。(3)本文将个性化推荐技术应用到智能决策推荐系统中。业务模型是智能推荐的基础和关键部分,直接影响推荐服务的优劣。从业务决策中抽取影响决策的属性,提取其信息结构,根据属性特点将业务和影响决策的属性组织为关系模型,再将模型映射为适合计算相似度的矩阵模型。本文在提出通用推荐模型的基础上,设计实现了一种对评估建议进行推荐的智能推荐系统,并在“国家免费孕前优生健康检查信息系统”中应用。本文的主要贡献是,提出了一种基于地点上下文的协同过滤推荐算法,通过实验验证了其对推荐系统准确度的提高。提出决策推荐的通用模型,并在“国家免费孕前优生健康检查信息系统”评估建议推荐功能中应用。更多还原

【Abstract】 With the popularity of Internet technology applications, the exponential growth trend in the number of flooding the network resources. The flood of information presented to the user at the same time, the "information explosion" and "information overload" phenomenon. The personalized recommendation system is an information service technology to alleviate this problem, it is based on user history and behavior information to build user interest model recommended by the information that may be of interest to the user through the model. Recommendation system on the one hand in the massive amounts of data by predicting user preference of the project to provide users with information filtering, application of knowledge discovery technology to generate personalized recommendation to help users find the information they need; other hand auxiliary enterprises achieve personalized marketing purpose, and thus increase sales, create more profits for the enterprise.Recommendation system with good development and application prospects has become an important research direction in Web intelligence technology, widespread concern by many researchers. In the past two decades, the personalized recommendation technology has been rapid development. With the advent of the era of big data, especially in the recommended system widely used in e-commerce, advertising push, there are still many problems to be solved in the study of the rapid growth of the music and movie recommendation data, personalized recommendation system. In this thesis, the key technologies for intelligent recommendation system user modeling and recommendation algorithm exploration and research. This thesis mainly personalized recommendation technology used in the recommendation system for enterprise decision-making. The main contents of this paper are as follows:(1) Similarity calculation does not consider some of the problems brought about by the user location context information based collaborative filtering recommendation system; this paper presents the design of a collaborative filtering algorithm based on location context information. The first according to the user’s location information of the system the calculation of the distance attenuation and the degree of similarity, and then based on the user interest in behavior between to build user preference relationship network; obtained through a combination of both, the degree of similarity between the users; Finally recommended for users.(2) The business model is the intelligent recommendation foundation and a key part of a direct impact on the pros and cons of the recommendation service. Attributes drawn from business decisions to influence decision-making, and to extract the information structure, the attribute organization for business and influence decision-making based on the attribute characteristics relational model to calculate the similarity matrix model, then the model is mapped to fit. On the basis of the proposed business model, this paper designed to achieve the recommended intelligent recommendation system, a decision-making and in the application of "national eugenics before pregnancy health check information system".更多还原