Sensor-Based Early Warning and Intelligent Traffic Diversion Framework for Landslide Risk Mitigation in Valley Regions Using Hybrid Machine Learning

Landslides in valley areas pose a considerable danger to road safety, particularly on steep slopes combined with heavy rainfall events. Monitoring systems tend to be reactive in nature, responding only once failure has been clearly visible for some time. This lack of close connection between monitoring systems and traffic management systems has greatly reduced the value of implementing preventative procedures. This research proposes an integrated sensor-based early warning system and intelligent diversion of traffic from hazardous conditions to provide rapid decision support through early identification and detection of hazards; as well as through the rapid and accurate management of traffic away from the hazard areas. The system relies on a framework of real-time monitoring and predictive analysis using multiple sensors such as vibration sensors for monitoring vibration patterns, tilt sensors for measuring slope angles, moisture sensors for recording moisture levels, and rain gauges for measuring rainfall conditions. Data resulting from these multiple sensors will be analysed with a hybrid machine learning approach; which consists of a Random Forest model combined with a Long Short-Term Memory (LSTM) model to classify risk areas. The accuracy of classification was 95.2%, the F1-score was 0.94, and the false alarm rate was 3.8%. Besides, the average detection latency was 1.4 seconds, which means that it can work almost in real time. Also, the diversion method has somewhat decreased the traffic jam by 41%, and the accident risk by 62.5% in the simulated environment. So, the main point of the paper is the combination of geotechnical sensing with intelligent transportation control, where the early warning message is used for automatic traffic diversion. To put it simply, it is a more feasible and extensive approach to enhancing road safety in valleys which are susceptible to landslides.