Methodology for mining prediction parameters based on network of nonlinear autoregressive moving average with exogenous factors

Abstract

The paper offers a methodology for forecasting the aerogas state of mining atmosphere with the use of artificial neural networks, autoregressive models and metaheuristics. It also suggests an improved AR model for forecasting the state of mine atmosphere by adding exogenous factors to its structure, which are measureable dynamic parameters of the gaseous state of mine workings. The metaheuristic algorithm is used to adapt the model. Numerical studies have shown that the proposed model can improve forecast accuracy by 10% as compared with the existing gradient methods