Uncertain decision tree for bank marketing classification

Abstract This study proposes a novel decision tree for uncertain data, called the uncertain decision tree (UDT), based on the uncertain genetic clustering algorithm (UGCA). UDT extends the decision tree to handle data with uncertain information, in which the uncertainty must be considered to obtain high quality results. In UDT, UGCA automatically searches for the proper number of branches of each node, based on the classification error rate and the classification time of UDT. Restated, UGCA reduces both the classification error rate and computing time and, then, optimizes the proposed UDT. Before the UDT is designed using UGCA, an uncertain merging algorithm (UMA) is also developed to reduce the uncertain data set, thereby allowing UGCA to process a large data set efficiently. Importantly, experimental results demonstrate that the proposed UDT outperforms traditional uncertain decision trees.