BankSegnet: a cloud-based framework for customer segmentation and term deposit prediction in retail banking

BankSegNet, a deep learning (DL)-based system for predicting customer subscriptions to bank term deposits within cloud-supported banking environments. The devised technique facilitates customer segmentation by securing access to data, data pre-processing and high-level feature learning. Initially, the term deposit dataset is fetched from the Google Cloud Storage and then subjected to imputation, one-hot encoding, and min-max normalization. After this, Weight of Evidence (WoE) transformation is used to map categorical attributes to the relative information value. The embedded categorical and numerical features are then passed through the one-dimensional convolutional neural network (1D-CNN) which learns feature interaction to segment bank customers and forecast term deposit subscriptions. Simulation results obtained shows that the BankSegNet achieves an accuracy of 98%, a precision of 98.81% and a recall of 97.96%.