Assessing factors influencing flood preparedness among Jakarta residents: A multilayer perceptron artificial neural network based on protection motivation theory

Abstract

As Greater Jakarta is highly vulnerable to flood disasters that pose significant economic risks, enhancing residents' preparedness has become a critical policy priority. This study extended the Protection Motivation Theory (PMT) by incorporating both external factors and PMT constructs to examine their influence on flood preparedness intentions. While PMT has been commonly applied using traditional methods like Structural Equation Modeling, such approaches often assume linear relationships and rely on rigid model structures. To overcome these limitations, this study integrated Artificial Neural Networks (ANN) as a flexible, data-driven method for identifying influential predictors. Following hyperparameter experimentation, the optimal Multilayer Perceptron model was configured with one hidden layer comprising forty nodes, using the Adam optimizer and Swish-Sigmoid activation functions for both hidden and output layers. The results showed that Coping Appraisal, Threat Appraisal, Flood Experience, Media Exposure, Geographical Perspective, and Government Action all positively influenced respondents’ Protection Motivation, with Coping Appraisal identified as the most influential factor. These findings provide actionable insights for local policymakers and organizations, while also demonstrating the potential of ANN as a powerful tool in behavioral disaster preparedness research.