Feature correlation measure based real time discrimination prevention with transactional data sets using social networks

Abstract

The effect of social network in anti-discrimination has been analyzed in detail.Various approaches towards anti-discrimination on transactional data have been identified, but does not produced expected performance level. To improve the performance, a feature correlation measure based approach has been presented in this article. The method reads the transactional data set and generates number of patterns based on the purchase details. For each pattern generated, the method estimate pattern impact measure towards the data set. Based on the value of PIM (Pattern Impact Measure), a subset of patterns is selected. With the selected pattern set, the method reads the social network user data and identifies the list of items being discussed. According to identified items, the method estimates the feature correlation measure (FCM) for each items identified. According to the value of FCM, a subset of items with higher value has been selected. The selected items are identified as more sensitive and being sanitized with the publishing data set. The sanitization is performed using probabilistic sanitization algorithm. The FCM algorithm has leverage the discrimination prevention and sanitization performance.