Classifying without discriminating

Abstract

Classification models usually make predictions on the basis of training data. If the training data is biased towards certain groups or classes of objects, e.g., there is racial discrimination towards black people, the learned model will also show discriminatory behavior towards that particular community. This partial attitude of the learned model may lead to biased outcomes when labeling future unlabeled data objects. Often, however, impartial classification results are desired or even required by law for future data objects in spite of having biased training data. In this paper, we tackle this problem by introducing a new classification scheme for learning unbiased models on biased training data. Our method is based on massaging the dataset by making the least intrusive modifications which lead to an unbiased dataset. On this modified dataset we then learn a non-discriminating classifier. The proposed method has been implemented and experimental results on a credit approval dataset show promising results: in all experiments our method is able to reduce the prejudicial behavior for future classification significantly without loosing too much predictive accuracy.

Data mining for discrimination discovery

“…The first one is to adapt the preprocessing approaches of data sanitization [Hintoglu et al 2005;Verykios et al 2004] and hierarchy-based generalization [Sweeney 2002;Wang et al 2005] from the privacy-preserving literature. Along this line, [Kamiran and Calders 2009] adopts a controlled distortion of the training set. The second one is to modify the classification learning algorithm (an in-processing approach), by integrating discrimination measures calculations within it.…”

Integrating induction and deduction for finding evidence of discrimination

“…The problem of discrimination preventing DSS, however, cannot be tackled without entering the details of the internal representation of the DSS. A first approach, dealing with data mining classifiers, is reported in [13]. Second, the approach based on classification rules could be extended to account for continuous decisions (e.g., wage amount, mortgage interest rate) and for continuous attributes (e.g., age, income) without resorting to apriori discretization.…”

Data Discrimination Prevention in Customer Relationship Managment

“…The privacy preservation documentation can adapt the hierarchy based generalization and data transformation which are the approaches of the preprocessing. For this [6] , [7] performs the disciplined exaggeration of the training data sets from which classifier can make minimum modifications in the data sets to get fair data. Preprocessing is mostly used when data mining is to be performed by external parties (other than data holders).…”