International Journal of
INTELLIGENT SYSTEMS AND APPLICATIONS IN
ENGINEERING
ISSN:2147-67992147-6799 www.ijisae.org Original Research Paper
Customer Churn Detection for insurance data using Blended Logistic
Regression Decision Tree Algorithm (BLRDT)
Nagaraju Jajam1*, Dr. Nagendra Panini Challa2
Submitted: 16/10/2022 Revised: 22/12/2022 Accepted: 05/01/2023
Abstract: Customer churn identification is indeed a subject in which machine learning has been used to forecast whether or not a client
will exit the corporation. Predicting client turnover is a difficult issue in a variety of businesses, with the financial sector being the most
well-known. Because of the constant updating of insurance plans, the client retention procedure is critical to the company's success.
Predicting client turnover and developing customer retention strategies are both key research areas in the insurance market. This identifies
the importance of detecting churn rate in the insurance industry. The primary objective of this research is to forecast consumer habits and
to distinguish between churners and non-churners during an earlier phase. We propose a Blended Logistic Regression Decision Tree
(BLRDT) framework for churn detection. Initially, the preprocessing of the insurance dataset is done by employing the Z-score technique.
Data splitting is done to separate the standardized dataset into training and testing sets. The churn prediction is done using the BLRDT
algorithm. The evaluation was done by using different measurements like accuracy, precision, recall, and f1-score and the outcomes are
depicted by employing the MATLAB environment. The research also discovered that the presented scheme seems to be a realistic and
slightly superior strategy for the insurance sector to estimate client turnover than the findings obtained using existing approaches.
Keywords: Customer churns detection, preprocessing, Z-score technique, and Blended Logistic Regression Decision Tree (BLRDT)
algorithm.
1. Introduction
The financial industry is one of the largest and most inefficiency and the inability to search for a specific
prominent industries that employs data forecasting technologies. item.Acquiring a new client, on the other hand, is able than keeping
Insurers' massive reliance on data and an ever-increasing customer an old one. Churn, defined as a consumer's willingness to
base is a major factor in this theory. Scholars and businesses alike discontinue a contract and go to another organization, has become
find this a daunting idea. Clients now have the option to transfer a big worry in several industries. The monthly ratio for American
insurance providers, which means the churn probability projection mobile networks is 3% to 4%, which represents a significant cost
in the insurance industry isn't meeting their demands. As a for corporations that pay 450 to 550 dollors to recruit a single client
consequence, churn predictions using ML techniques are who generates around 50 dollors in monthly income.Businesses
addressed. Because of its size and complexity, the financial are finally recognising the importance of service preservation.
industry undergoes rapid transformations in reaction to global According to one study, "the top six American telecom companies
economic shifts. An interesting study subject for students would still have gained 207 million dollars if they could have
interested in solving real-world challenges is the development of a retained an additional 5% of consumers who were ready to offer
trustworthy and honest customer relationship management system but changed contracts in the previous year."The corporation's main
(CRM). Insurance churn prediction is an important research area selling task in the coming decades is to reduce churn rates by
for fault detection and recognition operations. Customer churners detecting customers who are likely to depart and then making the
are on the rise because of technology that can't accurately predict required efforts to keep them. The first stage is to anticipate the
client turnover. Ullah, I et al. (2019) [1].According to marketing possibility of churn at the consumer level. There are 2 significant
experts, businesses lose 26 percent of their clients each year due to features of the Churn Forecasting model.
1* Research Scholar, School of Computer Science and
Engineering, VIT-AP • The first is that the number of churned clients (critical
University, Amaravati, Andhra Pradesh-522237, India. E- instances) is quite low (2% of all samples).
mail(s): nagaraju.jajam@vitap.ac.in • The second factor is precision. As a result, increasing
2Assistant Professor, School of Computer Science and monthly forecasting predictive performance by 1% could
Engineering, VIT-AP result in a $54 million increase in annual revenue for a
University, Amaravati, Andhra Pradesh, - 522237 India E- service with 1.5 million customers. Jamjoom, A.A., et al.
mail(s): nagendra.challa@vitap.ac.in (2021) [2].
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 78–89 | 78
