Predicting customer response to bank direct telemarketing campaign

Direct marketing enables banks and other financial organizations to focus on those customers who present the likelihood of subscribing to their products, offers, and other packages. Most often than not, identifying these group of customers poses a challenge to financial institutions. In line with the aforementioned, this study considered the typical case of bank direct marketing campaign dataset with two main objectives. First, to predict customer response to bank direct marketing by applying four classifiers namely, Multilayer Perceptron Neural Network (MLPNN), Decision Tree (C4.5), Logistic Regression and Random Forest (RF). Results from the study portray that Random Forest Classifier with an accuracy of 87% is the most prolific classifier in terms of predictive ability. Evaluation of the classifiers was performed using classification accuracy and ROC. The RF classifier produced 86.80% as well as 92.7% respectively to place first among the classifiers. The second objective was to identify the key features of customers who subscribed and are most likely to subsequently subscribe to term deposits. A cluster analysis was performed to achieve this goal. Results from the cluster analysis revealed that, customers on whom more call durations are spent on present a higher likelihood of subscribing to the term deposit. What is more, the results indicated further that customers with a minimum of secondary school education are good prospects to be targeted for the bank term deposit subscriptions.