Advanced Data Analytics and Machine Learning for Geospatial Interpolation, Anomaly Detection, Background Prediction and Emergency Evacuation Routing in Real-Time Radiation Monitoring System

For identifying natural trends, hotspots, hazardous areas, and mitigating potential health risk to the public and environment, spatial analysis of radiation levels is crucial. Machine learning models' implementation for prediction of background radiation levels and anomaly detection can bring a revolutionary change in radiation monitoring. Also, emergency evacuation routes with minimum radiation exposure are needed in times of radiological and nuclear emergencies. Traditional radiation monitoring systems lack advanced data analytics capabilities for comprehensive situational awareness and emergency response planning. Building on a companion paper that details a IoT connected radiation detector and monitoring website, this work fills the gap of the radiation monitoring analytics to map radiation via different interpolation techniques, anomaly detection, and evaluation of multiple machine learning models for background radiation prediction and emergency escape route planning using A* algorithm with minimum exposure to radiation in times of nuclear or radiological emergencies. The study setup involves using the mobile radiation detector from the companion paper and real data collection from around the Military Institute of Science and Technology (MIST), Dhaka, Bangladesh, and Institute of Nuclear Science and Technology (INST), Dhaka, Bangladesh. Collected data are visualized by plotting interpolated radiation heatmaps, anomaly detection, and background radiation prediction using machine learning and determining effective escape routes following the "as low as reasonably achievable" (ALARA) principle. To revolutionize the radiation monitoring system, simplifying the decision-making for the radiation protection professionals and public officials to develop radiation protection strategies has great potential.