Shikhi Agrawal and Rajesh Lavania , Dr. Bhimrao Ambedkar University, India

Mobile traffic prediction and call drop reduction remain critical challenges for telecommunications providers, as they directly affect user satisfaction and network efficiency. A call drop, defined as the unexpected termination of a call before either party ends the conversation, often results from network congestion, coverage deficiencies, or system malfunctions. Traditional statistical techniques have shown limited effectiveness in modeling the highly dynamic and complex nature of mobile networks. In contrast, recent research highlights the growing role of machine learning (ML) and data mining methodologies in capturing non-linear patterns, forecasting traffic loads, and improving quality of service (QoS). This review paper synthesizes ten recent studies (2021–2025) that explore models such as decision trees, random forests, gradient boosting, and deep learning architectures including LSTM and CNN-LSTM. Key findings emphasize the advantages of ML in enhancing prediction accuracy, optimizing network resources, and supporting energyefficient operations. However, persistent challenges such as data imbalance, feature selection, model interpretability, and scalability hinder real-time deployment. The review concludes by identifying open research directions, particularly the need for explainable, federated, and hybrid ML approaches that can be seamlessly integrated into 5G and future 6G networks.

Keywords