Optimizing Emergency Communication in Smart Cities using Random Forest Classifier

The enhancement of emergency communication systems in smart cities for mitigating road traffic accidents (RTAs) using machine learning (ML) techniques. It focuses on the application of supervised ML models, particularly Random Forest, integrated with feature selection methods like Principal Component Analysis (PCA) and Association Rule Mining (ARM), to classify and predict emergency scenarios based on the RTA dataset. The study achieves an overall accuracy of 80%, with PCA reducing training time by 30% and ARM improving interpretability by identifying significant accident-related patterns, such as correlations between road conditions, time of day, and severity. However, challenges like class imbalance evidenced by a single instance of Class 0 versus nine of Class 1 result in a precision-recall trade-off, with Class 0 showing perfect recall (1.00) but low precision (0.33). The paper highlights the potential of ML driven frameworks to optimize emergency response efficiency while addressing limitations such as data imbalance, scalability, and real-time processing. Future directions include integrating deep learning models (e.g., CNNs and RNNs), IoT-based real-time data acquisition, and federated learning to enhance predictive accuracy and applicability in dynamic urban environments, laying the groundwork for robust, adaptive emergency communication systems.