In the contemporary banking sector, direct marketing campaigns—specifically telemarketing—remain a primary vehicle for selling long-term deposit subscriptions. However, these campaigns often suffer from low conversion rates and high operational costs due to inefficient customer targeting. This study presents a systematic analysis of a direct marketing dataset from a Portuguese banking institution to identify key determinants of subscription success and predict customer behavior. Utilizing the KNIME Analytics Platform and IBM SPSS Statistics, we employed a multi-stage methodology encompassing rigorous data preprocessing, Exploratory Data Analysis (EDA), statistical hypothesis testing (T-tests, ANOVA), and predictive modeling (Logistic Regression, Decision Trees, and Multiple Linear Regression). Our results demonstrate that call duration is the most significant predictor of subscription likelihood, though it presents challenges regarding realistic implementation due to its ex-post nature. Furthermore, we identify that socio-economic factors such as education level and job category significantly influence account balances and subscription rates, while temporal factors like contact month exhibit strong seasonality. The predictive models revealed a dataset imbalance issue, with the Binary Logistic Regression achieving an overall accuracy of 89.1% but showing bias toward the majority class (non-subscribers). This paper discusses the implications of these findings for resource allocation and suggests strategies for handling class imbalance to enhance decision-support systems in banking.
Alhadi et al. (Sun,) studied this question.