一种保护原始数据的多属性值分类挖掘算法

针对基于随机响应的隐私保护分类挖掘算法仅适用于原始数据属性值是二元的问题,设计了一种适用于多属性值原始数据的隐私保护分类挖掘算法。算法分为两个部分:a)通过比较参数设定值和随机产生数之间的大小,决定是否改变原始数据的顺序,以实现对原始数据进行变换,从而起到保护数据隐私性的目的;b)通过求解信息增益比例的概率估计值,在伪装后的数据上构造决策树。

Algorithm for privacy-preserving classification mining with multivariate data

Randomized response technique was used in privacy-preserving classification mining,and had acquired good results.But the method was only fit for binary data. To solve this problem,this paper dasigned an algorithm which was fit for multivariate data for privacy-preserving classification mining.The algorithm divided into two parts.In the first part,compared the size of parameter and random generated number to decide whether the sequence of the original data should be changed or not,in order to disguise the original data and then protected the privacy of data.In the second part,estimated the value of gain ratio to build the decision tree on disguised data.