Advanced Sensor-Based Landslide Detection and Alert System Utilizing Machine Learning

Abstract

The accelerated frequency of landslides, worsened by climate changes, urbanization and natural geology, showcases the desperate need for suitably designed early warning programs. An “Advanced Sensor Based Landslide Detection and Alert System using Machine Learning” is designed here, which issues precise alerts which can impact the effects of the landslides. The system strategically isolates its machine learning and IoT components, thereby enhancing both hazard prediction and real- time tracking. A LightGBM model was adopted as the predictive model because of its speed and its property of being able to deal with categorical and numeric data effortlessly. The model is trained using extensive dataset obtained from DEM data with
15 landslide conditioning factors as the input variables. With these data, the model is able to calculate hazard probabilities with great accuracy mapping high-risk areas. Along with the predictive model, the high-risk areas marked by the ML model are supported by an IoT device. This set of devices contains three primary sensors: a rain sensor, a soil moisture sensor, and an ADXL-345 sensor which gives the accelerometer value and a GPS module guarantees precise geolocation. IoT sensors continuously track localized environmental data and send real- time information to a central cloud platform. The alerting mechanism of the system is implemented to have a low latency. As soon as a hazard is identified, warnings are sent directly from the cloud to a mobile app. The app not only warns the residents and the local authorities through push messages but also offers an interactive map that shows hazard areas and sensor update information. Integration with precise hazard mapping allows for prompt decision-making and timely emergency actions. The entire system constitutes an important advancement in disaster management technology. The isolation of machine learning and IoT features in the system enables high prediction accuracy as well as real-time responsiveness of operation, thus presenting a resilient and scalable option for protecting people living in areas exposed to landslides.