Threshold Based Real-Time Landslide Prediction System Using Low-Cost Sensor Networks

Landslides pose a recurring threat in the Himalayan region, leading to devastating consequences in terms of human casualties and property damage. This research introduces a groundbreaking approach to real-time landslide monitoring and prediction, leveraging an affordable sensor network. The system employs a network of wireless sensors strategically placed to continuously monitor environmental conditions and gather pertinent data. This data is then transmitted to a cloud server via a WiFi module, where it is integrated into the PowerBI platform for visualization and analysis. To predict landslides in real-time, the collected data is fed into a machine learning model. The performance of three algorithms, namely Support Vector Machine (SVM), K-Nearest Neighbor (K-NN), and Decision Tree (DT), was evaluated using the acquired data. Among these algorithms, SVM demonstrated exceptional results, exhibiting low values for key error metrics such as Mean Squared Error (MSE=O.067), Mean Absolute Error (MAE=O.246), and Root Mean Squared Error (RMSE=0.204). Threshold values are determined based on the training model's output, enabling the generation of alerts and warnings at different levels. This innovative system significantly enhances preparedness and response strategies in landslide-prone areas, equipping stakeholders with timely information to effectively mitigate the impact on lives and infrastructure.