流形学习及维数约简在数据隐私保护中的应用

采用流形学习及维数约简方法可以有效保护敏感数据。针对交通事故黑点的敏感数据挖掘中隐私保护问题，提出了综合应用等距变换和微分流形两种算法来提高原始数据保密程度的方法，采用基于旋转的等距变换扰乱数据，用Laplacian Eigenmap对高维数据进行非线性降维，在保留数据内在几何结构的同时，进一步扰乱数据。该方法有效地应用于交通事故黑点数据隐私保护中，同时降低了原始数据的维数，便于后续的数据挖掘与分析。

Application of manifold learning and nonlinear dimensionality reduction in private preserving

To deal with the privacy preserving problem during data mining for sensitive black points in traffic accidents,this paper presents a new method,which is based on the isometric transformation and the differential manifold,to improve the privacy preserving level of the original data.It disturbs the data by doing isometric transformation based on rotation.The nonlinear dimensionality reduction is done to high-dimensional data with Laplacian Eigenmap to further disturb the data,while preserving the inner structure of the data at the same time.This method is effectively applied to the privacy preserving problem during data mining for sensitive black points in traffic accidents,while reducing the dimensionality of the original data for later data mining and data analyzing.