PROJECT TITLE :

Multistep Fuzzy Bridged Refinement Domain Adaptation Algorithm and Its Application to Bank Failure Prediction

ABSTRACT:

Machine Learning plays an vital role in information classification and knowledge-based prediction. In some real-world applications, however, the training knowledge (coming from the source domain) and take a look at data (from the target domain) come back from different domains or time periods, and this could result in the different distributions of some options. Moreover, the values of the options and/or labels of the datasets would possibly be nonnumeric and involve obscure values. Traditional learning-primarily based prediction and classification strategies cannot handle these two issues. In this study, we tend to propose a multistep fuzzy bridged refinement domain adaptation algorithm, that offers an efficient manner to accommodate both problems. It utilizes an idea of similarity to change the labels of the target instances that were initially predicted by a shift-unaware model. It then refines the labels using instances that are most just like a given target instance. These instances are extracted from mixture domains composed of supply and target domains. The proposed algorithm is constructed on a basis of some data and refines the labels, therefore performing completely independently of the shift-unaware prediction model. The algorithm uses a fuzzy set-based approach to accommodate the obscure values of the options and labels. Four different datasets are used in the experiments to validate the proposed algorithm. The results, which are compared with those generated by the prevailing domain adaptation strategies, demonstrate a significant improvement in prediction accuracy in each the higher than-mentioned datasets.