A Novel Approach for Discrimination Prevention and Privacy Preservation in Data Mining

Abstract

Recently increasing importance of Data Mining technology as it will help for extract important knowledge data from large amount of data. Hence there negative sociality able to see about data mining. Peoples belonging some categories on that based peoples are treating unfairly. Data mining and data collection techniques classified the mining rules which is covered automated decisions, e.g. grant or denied loan request, insurance premium computation. Discriminatory attributes are gender, cast, region, race etc. if data set biased on above attributes decision may emanate. Discrimination having two types one is direct and indirect. When on the base of sensitive attributes made decision it’s called direct discrimination. When no any sensitive attribute include for made decision it’s called indirect discrimination and which are relate with biased sensitive one. In these studies we focus on discrimination prevention in data mining and our proposed system used for direct or indirect discrimination prevention individually or both at the same time. We focus on how to data discrimination decision convert in anti - discriminatory as cleaning training data set and outsourced dataset. Also we define the new metrics for evaluate with our approaches and we compare these approaches. This experiment explain that the proposed system how effectively removing direct or indirect discrimination while store data quality