Customer Churn Prediction Using Machine Learning Techniques: In the case of Abay Insurance

Insurance company has a vital role on the growth of national economy and industry. By providing an efficient and reliable insurance service for their clients, it is possible to get high revenue and increase the number of customers. But most business company including insurance sectors were affected by churning problem. Churn is the percentage of customers who have stopped using a products or service after a period of time. The cost of attracting new costumer can be five-ten times greater than the cost of retaining the existing one. For this reason, insurance companies pay more attention to their existing customer and attract them with more attractive and better offer and service to increase the level of the customer satisfaction. The purpose of this thesis work is to design and develop a churn prediction model using machine learning algorithm for Abay insurance to address churning problem. Predicting customer churn behavior provides valuable insights into the sustainable growth of a company by serving customers on a regular basis. A total dataset of 10400 with 10 attributes were used for analysis. After collecting the required data from Abay insurance in ‘Woldia branch’, a successive preprocessing steps like missing value handling, encoding, normalization, labeling, and feature selection are undertaken. By reviewing a numerous articles and nature of dataset, suitable machine learning models were selected. Each machine learning algorithms were analyzed using python version 3.9. Some of the algorithms used in this thesis work are Decision Tree, Support Vector Machine, Random Forest, Naïve Bayes, Logistic Regression and K-nearest Neighbors. The performance of each model is evaluated using performance metrics like accuracy, precision, recall, f1 score and AUC-ROC curve. Random Forest has an accuracy performance of 98.4% which is best from others and Decision Tree has 97. 78% the remaining classifiers SVM, NB, LR and KNN have an accuracy of 97.5%, 95.38%, 95% and 96.97% respectively. In AUC-ROC values, Random Forest has best among others which is 99% and Support Vector Machine is 98% and close to 1. Based on this reason, RF classifiers were selected for the proposed system. Generally, this study can address churning problem of Abay insurance by building churn prediction model using machine learning algorithm.<br>