Recurrent Neural Networks (RNN) and a Deep Learning (DL) approach based on Long-Short Term Memory (LSTM)

The intensification of key socioeconomic cycles’ indicators raises concerns about future trends and turning points of such indicators and their impact on economic and social growth and developments plans. The development of adequate early warning systems is crucial. Here, we used deep learning (DL) for deep socioeconomic parameter forecasting (such as urbanization rate, digitalization trends, defense and military industry localization and independence index, innovation ranking, privatization of national industries, Saudization rate of occupations, female participation rate in managerial positions, environmental parameters, natural resource utilization rate), along with more high frequency financial and economic variables to develop an effective forecasting system to provide timely information and support for decision making. For this purpose, we implemented a coupled model of a near-future global indicator forecasts with a short-range runoff forecasting system. Starting from a traditional time series conceptual models, we defined the risk indicators that were used in the data-driven unrestricted mixed-data sampling with ARMA component (UMIDAS-ARMA) forecast models. The deep parameter variables were obtained through statistical scaling of the global trends composite indicators, thus enabling two data-driven predicting approaches using a simple Artificial Neural Networks (ANN) and a Deep Learning (DL) approach based on Long-Short Term Memory (LSTM) Models. The coupling between the data-driven -runoff forecast models are able to satisfy two basic requirements that any early warning system should have: The forecast should be given in advance, and it should be accurate and reliable. In this context, DL significantly improves runoff forecast when compared with a traditional data-driven approach such as ANN, being accurate in predicting time-evolution of output variables, with an error of 5% for DL, measured in terms of the root mean square error (RMSE) for predicting the peak flow, compared to 15.5% error for ANN, which is adequate to warn communities at risk and initiate policy response recommendations.