High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

An Algorithm to Conserve Confidentiality Using Anti-Discrimination Method in Data Mining

Author(s):

Patel Shraddhaben Mukeshbhai , SWAMINARAYAN COLLEGE OF ENGINEERING AND TECHNOLOGY KALOL-382729 GUJRAT; Hiral Darji, ASSISTANCE PROFESSOR, DEPARTMENT OF COMPUTER ENGINEERING SCET

Keywords:

Anti-Discrimination, Data Mining, Direct and Indirect Discrimination Prevention, Rule Protection, Rule Generalization, Privacy, Discrimination Measures

Abstract

Data mining is increasing important technology. In Data mining, collection of large amount data is hidden. This technology is used for extracting useful knowledge hidden from large collection of data. The main two issues in data mining are privacy violation and discrimination. This technique is used in data mining for decision making in various classification. In real life observation, the majority people do not want to be discriminated base on their gender, nationality, religion, age and so on. This types of attributes is used for decision making purpose such as giving job, loan, insurance etc. so that discrimination issue is arise. For that reason to find such attributes and deleting them from training data without affecting their decision making utility important. Discrimination is two types’ direct discrimination and indirect discrimination. Direct discrimination occur when decision making utility is based on some sensitive attributes like race, religion, gender etc. Indirect discrimination occurs when decision making utility is based on non-sensitive attributes which are related to sensitive attributes. There are many new method propose for solving discrimination prevention problem by applying direct or indirect prevention one by one or both at same time. We discuss about how to clean training data sets and out sourced data sets in such way that direct and/or indirect discriminatory rules are converted to non-discriminatory rule. The propose system prevent the discrimination without affecting the data quality. In this paper, we mainly focus on anti-discrimination or hybrid approach (preferential sampling and direct and indirect discrimination) which helps to detect and prevent discrimination

Other Details

Paper ID: IJSRDV3I120069
Published in: Volume : 3, Issue : 12
Publication Date: 01/03/2016
Page(s): 475-478

Article Preview

Download Article