Supervised and unsupervised data mining approaches in loan default prediction

Given the paramount importance of data mining in organizations and possible contribution a data-driven customer classification recommender systems for loan-extending financial institutions, study applied supervised approaches to derive best classifier loan default. A total 900 instances with determined attributes class labels were used training cross-validation processes while prediction 100 new without labels. In phase, J48 confidence factor 50% attained highest accuracy (76.85%), <em>k</em>-nearest neighbors (<em>k</em>-NN) 3 (78.38%) IBk variants, naïve Bayes has 76.65%, logistic 77.31% accuracy. <em>k</em>-NN have accuracy, F-measures, kappa statistics. Implementation these algorithms test set yielded 48 non-defaulters 52 defaulters <em>k</em> -NN 44 56 under logistic. Implications discussed paper.