Abstract

In the telecom sector, which generates vast amounts of data daily due to its extensive client base, retaining present customers is more cost-effective than attaining new ones. Business analysts and CRM specialists must comprehend the reasons behind customer churn and identify behavioral patterns within client data. This study develops a churn forecast model employing clustering and classification algorithms to recognize churn consumers and highlight the factors influencing customer churn in the telecom industry. Feature selection is done using information gain and correlation feature ranking filters. The Random Forest (RF) algorithm achieved superior performance, correctly classifying 88.63% of churned customer data. An essential CRM function is to formulate effective retention strategies to prevent customer departure. Post-classification, the proposed model clusters the churned customer data and provides group-based retention strategies using cosine similarity among the groups. The performance of the model is assessed through metrics like precision, accuracy, recall, f-score, and ROC-AUC. The findings indicate that the RF algorithm enhances churn classification and customer profiling through k-means clustering, and the classification algorithm helps identify the factors driving customer churn through generated rules.

References

Citation

Copy

Copyright

Article metrics