Prevention Methods for Discrimination in Data Mining

Discrimination is a very important issue when considering as law and professional aspects of data mining. Data mining is an increasing crucial technology for extracting useful information hidden in large collections of data and negative social perceptions in data mining. This is providing along with privacy and security of the information and potential discrimination. The latter consists of unequal treating people depends on their belonging to a specific group of membership. Data mining methods used for data collection automating and data mining techniques that is classification rule mining have to cover the way to providing decisions automatically like loan granting or reject, insurance premium computation, etc. If the training data sets are cover in what regards discriminatory (sensitive) attributes like gender, race, religion, nationality etc. Discriminatory decisions may ensue because antidiscrimination techniques including discrimination discovery and prevention have been introduced in data mining. Discrimination can be providing either direct or indirect. Direct discrimination takes as when decisions are made depends on sensitive attributes. Indirect discrimination takes sa when decisions are made depends on nonsensitive attributes which are strongly relation with biased sensitive ones. In this paper handled by discrimination prevention in data mining and propose new techniques applicable for direct or indirect discrimination prevention individually or both at the same time and how to clean training data sets are handle and outsourced data sets in used to cover way to that direct and/or indirect discriminatory decision rules are converted to perfect (nondiscriminatory) classification rules and providing loan is process is speed. the important advantages for manager how because he is easily check details and grating loan. The practical result demonstrate that the proposed methods are effective at removing direct and/or indirect discrimination biases in the original data set with preserving data quality.