Using novel noise filtering techniques to ensemble classifiers for credit scoring: Using noise filtering techniques to ensemble classifiers for credit scoring

Credit scoring using ensemble classifications has been applied widely. Ensemble classifications benefit from each single models, however biased models and miss-classification happens because of noisy data. To improve the performance of ensemble classifications, the proposed model extends the noise filtering technique that increases the accuracy of the deployed classifier. In addition, a new hybrid model combines novel noise filtering techniques and ensemble classification methods for scoring based on voting. In the hybrid model, the new label noise of data is assigned by support vector machine, then ensemble classifications to score noise. Using the novel noise filtering techniques and ensemble classifications for credit scoring obtains the high robustness and the superior performance. To be fair, some data sets for credit scoring problem are executed to compare results. The performance of the proposed model implemented by the experiment is better than the other models from 1% to 2%.