基于隐私保护的社交网络用户推荐系统

发布时间：2018-05-18 20:02

本文选题：推荐系统 + 轨迹隐私　；参考：《西南科技大学》2017年硕士论文

【摘要】：随着移动社交网络的快速发展,基于GPS位置的服务给人们的生活带来了很多的乐趣与便捷。然而在享受LBS便捷服务的同时,用户的位置隐私保护也面临越来越大的考验,如何保护位置隐私成为了研究的热门话题。本文研究了现有位置隐私保护算法的特点和不足,在k-匿名算法和泛化法的基础上提出了基于位置坐标均值的隐匿算法。算法取用真实位置均值坐标生成隐匿区域,并把整个区域作为用户的位置发送给应用服务器;应用服务器通过计算隐匿区域的圆心生成推荐区域,并且根据推荐区域的商家信息为用户提供推荐服务。这种算法在隐私保护度、匿名成本、匿名时间等性能上有很大的优势,而且在给用户提供优质位置隐私保护的同时不影响推荐效果。应用服务端在提供服务时获取了大量的用户位置信息,对这些用户基本信息和用户历史记录加以利用,能够给用户提供基于位置的个性化推荐。本文通过爬取大众点评的商家数据,通过逻辑回归算法对数据模型训练获得模型结果。根据用户当前位置求取区域范围内的商家信息,利用逻辑回归和协同过滤算法,成功实现了基于位置隐私保护的个性化商家推荐系统,并且实现的系统在推荐准确率上很有优势。互联网数据呈现爆炸式增长,解决信息超载问题成为一大难题,传统处理大数据方法优势显得越来越弱,因此本文采用处理数据效率更高的大数据集群运算框架Spark实现,通过机器学习库Spark MLlib完成了基于隐私保护的推荐系统设计。
[Abstract]:With the rapid development of mobile social networks, GPS location-based services bring people a lot of fun and convenience. However, while enjoying the convenient services of LBS, the location privacy protection of users is also facing more and more challenges. How to protect location privacy has become a hot topic. In this paper, the characteristics and shortcomings of the existing location privacy protection algorithms are studied. Based on the k-anonymity algorithm and the generalization method, the hiding algorithm based on the location coordinate mean is proposed. The algorithm uses the real location mean coordinates to generate the hidden area, and sends the whole region to the application server as the user's position. The application server generates the recommendation area by calculating the center of the hidden region. And provides the recommendation service for the user according to the merchant information of the recommendation area. This algorithm has great advantages in privacy protection, anonymity cost, anonymous time and so on, and it does not affect the recommendation effect while providing users with high quality location privacy protection. The application server obtains a large amount of user location information when it provides the service, and makes use of these users' basic information and user history records, which can provide users with personalized recommendation based on location. In this paper, the model results are obtained by crawling the merchant data of Dianping and training the data model by logical regression algorithm. According to the current location of the user to obtain the business information within the scope of the region, using logical regression and collaborative filtering algorithm, the personalized merchant recommendation system based on location privacy protection is successfully implemented. And the implementation of the system in the recommendation accuracy is very good. With the explosive growth of Internet data, solving the problem of information overload has become a big problem, and the advantage of traditional big data processing method is becoming weaker and weaker. Therefore, this paper uses the big data cluster operation framework Spark which is more efficient to process data. The recommendation system based on privacy protection is designed by machine learning library Spark MLlib.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【相似文献】