The percentage of consumers or subscribers that discontinue using a product or service within a given time frame is known as the “churn rate.” Hence, using machine learning models to estimate the number of possible churn consumers is crucial for businesses to retain customers and enhance profitability by identifying at-risk customers early. With the increasing availability of customer data and advancements in machine learning techniques, accurate churn prediction has become more feasible and impactful. This research compares and analyzes the advantages and disadvantages of three different machine learning algorithms applied to customer churn prediction: random forest, decision tree, and neural network. The results demonstrate the superiority of the neural network when it handles unstructured and high-dimensional data compared to the decision tree model due to its higher F-measure score of 77.48%. The ability of the decision tree model to capture complex, non-linear relationships among attributes is limited. However, in uncomplex customer churn predictions, the decision tree model gets a high prediction score due to its accuracy rate of 90.8%. Yet, the random forest model works superior because to its 91% accuracy rate.
Xiaolei Jia (Thu,) studied this question.