Chaos-generalized regression neural network prediction model of mine water inflow

Abstract Artificial neural network (ANN) provides a new way for mine water inflow prediction. However, the effectiveness of prediction using ANN model would not be guaranteed if influencing factors are difficult to quantify or there only few observation data. Chaos theory can recover rich dynamic information hidden in time series. By reconstructing series phase space, multi-dimensional matrix could obtained, with each column representing an factor and its value change time. Therefore, established by combining chaos when lacking data on inflow. In present study, No. 12 coal Pingdingshan China was selected as study site. The Chaos-GRNN Chaos- BPNN mine, were from February 1976 December 2013. verified values 24 months 2014 2015. number embedded dimension ( M ) determined space reconstruction 7, meaning that 7 neurons GRNN input layer, delay 13 months. layer accordingly. maximum Lyapunov index 0.0530, 19 two models evaluated four evaluation indices R , RMSE, MAPE, NSE) violin plot. It found both realize long-term inflow, is better than Chaos-BPNN model.