Toward Privacy-Preserving Personalized Recommendation Services

Recommendation systems are crucially important for the delivery of personalized services to users. With personalized recommendation services, users can enjoy a variety of targeted recommendations such as movies, books, ads, restaurants, and more. In addition, personalized recommendation services have become extremely effective revenue drivers for online business. Despite the great benefits, deploying personalized recommendation services typically requires the collection of users’ personal data for processing and analytics, which undesirably makes users susceptible to serious privacy violation issues. Therefore, it is of paramount importance to develop practical privacy-preserving techniques to maintain the intelligence of personalized recommendation services while respecting user privacy. In this paper, we provide a comprehensive survey of the literature related to personalized recommendation services with privacy protection. We present the general architecture of personalized recommendation systems, the privacy issues therein, and existing works that focus on privacy-preserving personalized recommendation services. We classify the existing works according to their underlying techniques for personalized recommendation and privacy protection, and thoroughly discuss and compare their merits and demerits, especially in terms of privacy and recommendation accuracy. We also identity some future research directions.     

基于用户浏览轨迹的商品推荐

随着电子商务的迅速发展,推荐系统在这些网站中得到了广泛的应用。目前应用最广泛的个性化推荐算法是协同过滤推荐算法,但是该方法存在稀疏矩阵与冷启动问题。根据用户浏览记录推荐商品是缓解这些问题的一个重要研究方向,这些方法根据用户在电子商务网站的访问日志,提取出用户的浏览路径序列,即用户浏览轨迹,为用户推荐偏爱商品。目前,通过分析用户浏览路径为用户推荐商品的方法主要依据用户浏览轨迹模式匹配或者从用户浏览轨迹中商品与下一个商品关系的角度进行考虑。而本研究从浏览轨迹中被浏览商品与最终被购买商品关系的角度出发,并以此为基础建立用户浏览轨迹偏爱模型,挖掘用户偏爱,为用户推荐商品。实验表明,所提方法能够在一定程度上解决因为新用户缺少历史购买及评分记录而引起的新用户冷启动问题,提高了推荐方法的准确度与召回率。     

考虑DG接入影响配电网经济运行时间的线损分摊模型

每一区域配电网都有使其网损达到最小值的经济运行区间,而分布式电源(DG)的接入会影响配电网的运行状况,有可能使配电网偏离或进入经济运行区间。为了维持配电网的经济运转,文中提出了一种考虑DG接入影响配电网经济运行时间的线损分摊方法。首先,基于等值电阻法计算配电网线损,以配电网线损最小为目标,引入配电网经济运行时间的概念,进而计算得到DG接入前后该时段配电网所处的负载区间;其次,通过分析DG接入前后配电网负载区间发生变化的原因,形成不同的DG并网运行场景;然后,在潮流追踪分摊网损的基础上,为不同运行场景下的DG分别设置不同的奖惩系数,修正DG线损分摊量;最后,以某地区实际运行的线路为例进行计算分析,并与潮流追踪算法计算的结果比较,验证所提方法的正确性和有效性。     

一种个性化推荐方法

提出了一种新的个性化推荐方法。该方法来源于对个性化推荐技术本质的研究。产出的方法包括一种用正态分布卷积性质所得到的离线相似度计算方法;一种通过计算物品与物品之间无差别的相似性操作次数得到离线相似度的方法;一种用类似于贝叶斯的方法来综合不同的相似度结果的方法。另外还提到一些用于工程实施的方法和技巧。所提方法已经在数据挖掘领域得到了成功的应用。     

浅谈个性化信息服务

网络环境下,信息的海量增长和分布无序与用户日益增长的个性化信息需求存在着尖锐的矛盾,传统模式下的信息服务难以解决这个矛盾,更难以满足不同用户的特定需求。在此背景下,个性化信息服务应运而生,走上了信息服务发展的道路。本文从分析个性化信息服务的基本内涵入手,探讨了开展个性化信息服务的意义,阐述了个性化信息服务的特点和服务方式。     

基于微信平台的图书个性化借阅服务研究

本文通过对基于微信平台开展图书个性化借阅服务的必要性和可行性进行分析,提出了基于微信平台的图书个性化借阅服务的各项功能设计,并指出了服务开展中应关注意的相关事项,以期能够更好地促进当前图书馆图书借阅服务的创新发展.     

移动学习在教学过程中的应用

随着移动计算技术的发展,具有移动性、及时性、交互性等特点的移动学习技术得到了越来越广泛的关注和应用。为达到教与学过程监测和个性化指导的目的,在教学过程中使用移动学习技术辅助教学,学生通过手机浏览器访问课程资源、测试、作业和问卷,教师根据量化、统计、分析学生对知识点的掌握情况监测学生学习并及时给予个性化指导。结合目前教学的实际情况,探讨移动学习在教学过程中的应用模式和实际效果。     

用户兴趣模型的更新与遗忘机制研究

根据HTML文档的特点,赋予不同标签限定的特征词相应的权重系数,从而计算特征词的权值,并根据用户的浏览速度更新兴趣模型,其次,采用在在用户模型的特征向量中引入该特征词最后一次的更新时间,并结合遗忘因子修正特征词的权重的方法,实现了模型的遗忘。实验证明,引入遗忘机制的个性化推荐系统获得了较高的推荐效率。     

Video-based personalized traffic learning

We present a video-based approach to learn the specific driving characteristics of drivers in the video for advanced traffic control. Each vehicle’s specific driving characteristics are calculated with an offline learning process. Given each vehicle’s initial status and the personalized parameters as input, our approach can vividly reproduce the traffic flow in the sample video with a high accuracy. The learned characteristics can also be applied to any agent-based traffic simulation systems. We then introduce a new traffic animation method that attempts to animate each vehicle with its real driving habits and show its adaptation to the surrounding traffic situation. Our results are compared to existing traffic animation methods to demonstrate the effectiveness of our presented approach.     

Personalized mode transductive spanning SVM classification tree

Personalized transductive learning (PTL) builds a unique local model for classification of individual test samples and is therefore practically neighborhood dependant; i.e. a specific model is built in a subspace spanned by a set of samples adjacent to the test sample. While existing PTL methods usually define the neighborhood by a predefined (dis)similarity measure, this paper introduces a new concept of a knowledgeable neighborhood and a transductive Support Vector Machine (SVM) classification tree (t-SVMT) for PTL. The neighborhood of a test sample is constructed over the classification knowledge modelled by regional SVMs, and a set of such SVMs adjacent to the test sample is systematically aggregated into a t-SVMT. Compared to a regular SVM and other SVMTs, a t-SVMT, by virtue of the aggregation of SVMs, has an inherent superiority in classifying class-imbalanced datasets. The t-SVMT has also solved the over-fitting problem of all previous SVMTs since it aggregates neighborhood knowledge and thus significantly reduces the size of the SVM tree. The properties of the t-SVMT are evaluated through experiments on a synthetic dataset, eight bench-mark cancer diagnosis datasets, as well as a case study of face membership authentication.