Improving the Accuracy of Predicting Bank Depositor’s Behavior Using a Decision Tree

Abstract

Telemarketing is a widely adopted direct marketing technique in banks. Since customers hardly respond positively, data prediction models can help in selecting the most likely prospective customers. We aim to develop a classifier accuracy to predict which customer will subscribe to a long-term deposit proposed by a bank. Accordingly, this paper focuses on a combination of resampling, in order to reduce the imbalanced data, using feature selection, to reduce the complexity of data computing and dimension reduction of inefficiency data modeling. The performed operation has shown an improvement in the performance of the classification algorithm in terms of accuracy. The experimental results were run on a real bank dataset and the J48 decision tree achieved 94.39% accuracy prediction, with 0.975 sensitivity and 0.709 specificity, showing better results when compared to other approaches reported in the existing literature, such as logistic regression (91.79 accuracy; 0.975 sensitivity; 0.495 specificity) and Naive Bayes classifier (90.82% accuracy; 0.961 sensitivity; 0.507 specificity). Furthermore, our resampling and feature selection approach resulted in improved accuracy (94.39%) when compared to a state-of-the-art approach based on a fuzzy algorithm (92.89%).

Modeling the Telemarketing Process Using Genetic Algorithms and Extreme Boosting: Feature Selection and Cost-Sensitive Analytical Approach

“…However, due to the imbalance distribution of the data and other characteristics, non of the related works was able to provide solutions that handle all the challenges faced in modeling this business issue [5], [7], [19]. Interestingly, most of the researchers agree that not all the dataset features have the same importance in understanding the intentions of the clients [4], [5], [7], [9], [10], [20]. Table 1 describes, in brief, some of the related works and the best-attained prediction performance in terms of Geometric Mean (GMean) and Type I Error.…”

“…Table 1 describes, in brief, some of the related works and the best-attained prediction performance in terms of Geometric Mean (GMean) and Type I Error. Some of the works listed in Table 1 reported the modeling performance of different approaches and Machine Learning (ML) algorithms [4], [6], [21] and some attempted to understand the significance of the specific dataset features in predicting the willingness of the client to accept an offer [5], [9], [10].…”

“…Safarkhani, and Moro [9] proposed a combinatory approach that does feature selection and handles imbalanced data. They utilize wrapped subset evaluation to select the best discriminating features.…”

“…While the proposed solution performs well in terms of GMean, it made quite poor performance in terms of type I error. In addition to this performance flaw, the way [9] handles imbalanced data is questionable, as both oversampling and under-sampling are maintained simultaneously. such that correctly classified instances were increased and wrongly classified instances were reduced.…”

“…In practice, researchers aim at leveraging the marketing process success rate either by implementing highly demanding and complex algorithms [5], [6], [8] (e.g., Deep Learning and Artificial Neural Networks) or employing approaches in which the original data are manipulated [4], [9]- [12] (e.g., resampling). Complex algorithms may generate competitive models but hinder the interpretation of the outcomes by decision-makers, and data manipulation may impose computational overhead or a distorted real-world representation of the problem.…”