A predictive model for weather-related risks utilizing factor analysis and long short-term memory networks

This study introduces a holistic approach to outcome prediction using multiple indicators. We develop an innovative methodology that combines factor analysis for weight determination of variables, facilitating dimension reduction, with Long Short-Term Memory (LSTM) neural networks for comprehensive evaluation and precise forecasts. To assess the efficacy of our proposed technique, we undertook empirical research in the domain of weather-related risk—a field of growing importance that demands further investigation. The findings indicate that our predictive model, which merges factor analysis with LSTM, excels in multivariate forecasting, yielding relatively consistent results. Moreover, it underscores the projection that the impact of weather-related disasters is likely to escalate in tandem with population and economic growth.