Mining negative links between data clusters

Abstract

Link discovery (LD) is an important task in data mining for identifying interactions between data groups, or relating in society community networks. A new strategy is designed for mining a new kind of link: negative links between data clusters. The efficiency is gained by pruning strong positive relative items. Negative item is computing with correlation coefficient. The number of the negative item correlation is used to identify the negative links between clusters. These negative links are extremely useful in business fraud, medical treatment and incursion detection. Experiments on real datasets illustrate that our approach is efficient and promising.