Background: In recent time with the growth of the technology and the business model,
customer attrition analysis is considered as a very important metric which decides the revenues and
profitability of the organization. It is applicable for all the business domains irrespective of the size
of the business even including the start-ups. Because about 65 % revenue for the organization
comes from the existing customer. The goal of the customer attrition analysis is to predict the customer
who is likely to exit or churn from the current business organization. In this research work,
the literature review is carried out to explore the related work which has been already carried out in
the field of customer attrition analysis. The literature review also focuses on some of the patents
which are issued in the area of customer attrition or churn analysis. The goal of the research paper
is to predict accurately the customer attrition rate in the Banking Sector.
Objective: The main objective of this paper is to predict accurately the attrition rate in the Banking
sector using an optimized deep feed-forward neural network.
Method: In the proposed work the predictive machine learning model is implemented using the optimized
deep feed-forward neural network having five hidden layers in it. The model is trained using
Adam optimizer algorithm to obtain the optimal accuracy. The Banking Churn data set is
passed as input to the Optimized Deep Feed Forward Neural Network Model. In order to perform
the comparative analysis, the same data set is passed as input to the other machine learning algorithm
such as Decision Tree, Logistic Regression, Gaussian Naïve Bayes, and Artificial Neural
Results: The test results indicate that the proposed optimized deep feedforward neural Network
model performed better in accuracy compared to existing machine learning techniques.
Conclusion: The proposed optimized deep neural network model is an accurate model for customer
attrition analysis in the Banking sector compared to the existing machine learning techniques.