Algorithm 1. Overview of the proposed landslide-risk prediction algorithm

(1) Build the landslide dataset by using QGIS: (1.1) generate grid-based maps of 3 historical landslide occurs of Study areas given by Aerial photographs, Google Satellite, and site survey photographs. (1.2) identify 5 types of land cover from Google Satellite and Aerial photographs and render land cover digital map. (1.3) determine physiographic by using Aerial photographs and site survey photographs. (1.4) compute and render elevation and slope from Aerial photographs and site survey photographs. (1.5) fetch historical rainfall data from Department of Water Resources, Ministry of Natural Resources and Environment, Thailand. (1.6) export all attributes into the CSV files. (2) Determine the optimal values of batch size, epochs, and the number of nodes in full-connected network of ANN, GRU, LSTM, and Bi-LSTM. (3) Build and train the machine learning model i.e., LR, ANN, GRU, LSTM, and Bi-LSTM by using the optimal values of parameter from Step 2 given by the landslide dataset. The landslide dataset was split into a training dataset (80% of landslide dataset) and a testing dataset (20% of the landslide dataset) (4) Choose the model that delivers the best prediction performance for constructing the web application. (5) Implement and test the automatic landslide-risk prediction web GIS by using JavaScript, Node.js, and MySQL as web programing language, web development tool on server side, and database. (6) Deploy the proposed landslide-risk prediction web GIS on Google Cloud platform