Smart Campaign for Smart Business Using Machine Learning on Improving Product Marketing Strategy

Abstract

Improving the effectiveness of banking marketing campaigns is a major challenge in retaining and attracting new customers. This research aims to predict the success of direct marketing campaigns using the Random Forest algorithm on the UCI Bank Marketing dataset. The dataset includes various demographic variables and historical customer interactions. Preprocessing was done through min-max normalization and division of training and test data with a ratio of 80:20. Model evaluation results show that Random Forest is able to classify uninterested customers with high accuracy (true negative = 794), but has a weakness in detecting interested customers (true positive = 9), indicating a class imbalance in the data. The overall accuracy of the model reached 87%, with a precision of 64% and a recall of 8.7%. Feature importance analysis showed that the variables balance, age, and day were the most influential factors in customer decisions. Overall, the Random Forest algorithm successfully uncovered important patterns in the data that are relevant for more targeted marketing decisions. Nonetheless, improving the model's performance towards minority classes needs to be done through the approach of handling data imbalance. This research contributes to the utilization of machine learning in supporting data-driven marketing strategies in the banking sector.

Keywords: Random Forest; technopreneur; smart campaign