Abstract

Customer churn in the telecommunications industry is a problem of great concern to companies. The high cost involved in acquiring new customers has shifted the telecoms sector's focus on retaining new customers. With an average churn rate of 1.9%, the ability to predict accurately when a customer might churn has become an asset to telecommunications companies. This research aims to find a subtle method of predicting customer churn by applying machine learning algorithms - Decision Tree, Logistic Regression, and Support Vector Machine – to a dataset and evaluate how they stack up against each other in determining customer churn. The dataset provided for this research comes from Amplia Communications Limited, a major ISP in Trinidad and Tobago. Python is used to develop the IT artifact in the research. The dataset is pre-process using label-encoding, one-hot encoding, and Pearson correlation is used for feature selection. The processed dataset is then normalized and split into 70%/30% for training and testing, respectively. Using the evaluation metrics of Accuracy, AUC, and F1 score, the Decision Tree model with values of 89.5%, 81.9%, and 86.9% respectively, outperformed the other two machine learning models. In conclusion, the decision tree model would be the best suited for predicting customer churn from the results.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call