Discriminative Pattern Identification using Rule Based Approach.

Discrimination is bias behavior of people in society; particularly discrimination is based on race, sex, age and cast. Discrimination observed in many areas like labour market, education credit, mortgage and medical. Most of scientist found it in many subject like social sciences economics and law. Discrimination system relies on historical data for making decisions in socially sensitive actions. The technique of Discrimination identification uses information systems based on data mining technology for decision making. Decision making systems and data mining techniques such as association rule mining have been designed and are now used for making automated decisions, like loan granting or denial. Discrimination situations found in dataset in direct and indirect ways. Rules are formed from the dataset using Apriori algorithm and certain parameters such as number of rules, minimum support and confidence. Power of discrimination within rules are calculated by an elift and glift on a classification rule using Alpha and strong Alpha protection. In direct discrimination, the rules are directly extracted from dataset and searched for discriminatory pattern. In indirect discrimination, system needs some background knowledge as a further input and it is used to find unfair treatments. Inference model required for integration of classification rules with background rules. Inference model is a mathematic model. Direct and Indirect Discrimination tested over German credit dataset.