Abstract

Churn is characterized to be the movement of customers leaving the organization and disposing of the administrations offered by
it because of the disappointment of the administrations as well as better offering from other network suppliers. To carry out a
comparative analysis of the existing churn management models, we study the various characteristics of existing models based
on techniques used methods of data classification and feature selection processes. Based on this comparison, this study can
discover various types of knowledge, including association, classification, clustering, prediction, sequential patterns and decision
tree. The knowledge acquired from this comparison will then be classified into general knowledge, primitive-level knowledge, and
multilevel knowledge. To model the Customer prediction, a Markov Chain Model will be used. The Markov model allows for more
flexibility than most other potential models, and can incorporate variables such as non-constant retention rate, which is not possible
in the simpler models. The model allows looking at individual customer relationships as well as averages, and its probabilistic
nature makes the uncertainty apprehensible. The purpose of thisstudy was to ascertain the relevant drivers of customers` churn
and retention in the growing telecommunication industry especially in Nigeria and developed an enhanced predictive model to
address earlier limitation of accuracy and improved churn prediction. The enhanced churn prediction model performed better
than the unenhanced model. Logistic regression had better performance metric than other algorithms: neural network, Support
vector machine, decision tree and random forest. Although, all the other algorithm had a high AUC but in terms of generality
and simplicity logistic regression resulted in the highest AUC value on performance statistics – Accuracy, Sensitivity, Specificity.
More so, the result showed that internet service, types of contract entered, internet security were major factors that influence
churn.