Integrated Flood Mapping and Pattern Recognition: Leveraging Geographic Information Systems (GIS) for Risk Assessment

This research aims to develop a flood prediction model using Geographic Information System (GIS) technology and machine learning algorithms to predict daily flood events at the administrative area level in Jakarta. The goal is to improve flood risk assessment by integrating pattern recognition with GIS, thus enabling stakeholders to make informed decisions for flood mitigation. Based on data from BPBD Jakarta and Jakartaone, our model achieved test accuracy between 85% and 90% across various administrative areas. This accuracy range is calculated based on the overall accuracy of the predictions against the actual data, but to provide a more comprehensive picture, metrics such as precision, recall, and F1 score are also used. Precision measures the proportion of correct positive predictions, recall assesses how well the model detects flood events, and F1 score is the harmonic mean between precision and recall, which is ideal for imbalanced class distributions. The model was developed using the Random Forest algorithm, which is effective in handling data with many features and non-linear relationships. Once flood data and other predictive variables (such as rainfall, elevation, and land use) are collected and processed, the data is divided into training and test sets, with validation performed using cross-validation methods to avoid overfitting. The final results are visualized through ArcGIS, providing thematic maps and webGIS dashboards to show vulnerable areas and potential flood events. This integration of GIS and machine learning in disaster management is a significant step in strengthening flood preparedness and response, while reducing social and economic impacts in urban areas.