Integrated Churn Prediction and Customer Segmentation Framework for Telco Business

In the telco industry, attracting new customers is no longer a good strategy since cost of retaining existing much lower. Churn management becomes instrumental in industry. As there limited study combining churn prediction and customer segmentation, this paper aims to propose an integrated analytics framework for management. There are six components framework, including data pre-processing, exploratory analysis (EDA), prediction, factor analysis, behaviour analytics. This integrates segmentation process provide operators with complete better manage churn. Three datasets used experiments machine learning classifiers. First, status predicted using multiple Synthetic Minority Oversampling Technique (SMOTE) applied training set deal problems imbalanced datasets. The 10-fold cross-validation assess models. Accuracy F1-score model evaluation. considered be important metric measure models premise able identify who will Experimental indicates that AdaBoost performed best Dataset 1, accuracy 77.19% 63.11%. Random Forest 2, 93.6% 77.20%. 3 terms accuracy, at 63.09%, while Multi-layer Perceptron F1-score, 42.84%. After implementing Bayesian Logistic Regression conduct figure out some features segmentation. then carried K-means clustering. Customers segmented into different groups, which allows marketers decision makers adopt retention strategies more precisely.