Application of a Dynamic Recurrent Neural Network in Spatio-Temporal Forecasting

Spatio-temporal data mining is the extraction of unknown and implicit knowledge, structures, spatio-temporal relationships, or patterns not explicitly stored in spatio-temporal databases. As one of data mining techniques, forecasting is widely used to predict the unknown future based upon the patterns hidden in the current and past data. In order to achieve spatio-temporal forecasting, some mature analysis tools, e.g., time series or spatial statistics, are extended to spatial or temporal aspect, respectively. Among other methods, neural network is widely used for spatial forecasting. Normally a static forward neural network is employed to discover the hidden and deeply entangled spatial relationships. However, such approach is insufficient in forecasting dynamic process developing over space (such as forest fire). Elman is a kind of dynamic recurrent neural network (RNN) which allows the network to detect and generate time-varying patterns as well as spatial-varying patterns. Therefore, we use the Elman network for spatio-temporal forecasting. Experimental results collected from real case of forest fire area prediction confirm the viability and effectiveness of the proposed methodology.