Vol. 6, Issue 2, Part A (2024)

This research endeavors to enhance telecom airtime credit risk prediction through the application of machine learning algorithms. For financial stability and customer satisfaction, ethio telecom, the top telecommunications provider in Ethiopia, must effectively manage credit risk. Accurate credit risk pre- diction can assist the business in identifying clients who are more likely to default on their airtime credit, enabling proactive measures to reduce risks and improve financial performance. The historical customer information included in the dataset for this study includes customer profiles, call records, credit repayment histories, and usage data. Data preprocessing techniques are used before model training to handle missing values, encode categorical variables, and reduce features, ensuring the quality and consistency of the dataset. Machine learning algorithms such as Random Forest, Logistic Regression, Na¨ıve Bayes, and K- Nearest Neighbors (KNN) are leveraged to construct a predictive model under different experimental conditions. After controlling the impact of class imbalance and introducing novel attributes, experimental result shows that Random Forest and Logistic Regression machine learning algorithms exhibit promising results in predicting airtime credit risk. One of the major challenges in this research is dealing with class imbalance in the dataset, where the number of instances of customers who default on their airtime credit is significantly higher than those who do not. To address this challenge, future work should focus on implementing advanced techniques for handling class imbalance, such as synthetic data generation (e.g., SMOTE) and exploring ensemble methods that combine multiple algorithms to improve predictive performance. Additionally, continuously incorporating new and relevant attributes and refining the feature selection process will further enhance the model’s accuracy and reliability.