A New Method of Unsupervised Link Discovery Based on the Relative Density
Wu Shan, Zhiwei Ni, He Luo, Zheng Ying-ying · Zhongguo guanli kexue · 2008
Using K-neighbor(KNN)algorithm to solve novel node discovery problems usually has certain limitations and deviations during the process of data mining.The paper presents the concepts of the weighted distance and the relative density according to the above problems,and measures the local outlier degree of an object by its relative density based on the weighted distance.On this basis,the paper suggests a new method of unsupervised link discovery based on the relative density.The experiment shows that the new method has a better precision and can describe the outlier degree of the object more accurately.