The factors influenced infrared radiation drying rates for Agaricus bisporus, such as radiation intensity, radiation distance, material temperature, material thickness and drying time were analyzed. The network model structure between moisture content and all the controlling factors was built based on feed-forward neural network, the selected structure of the applied neural network, with its five inputs, single output and 11 hidden neurons were used. All data series obtained from different drying runs were used for training and test, mathematical model responding to inner relationship of the experimental data was obtained by finite iteration calculation, and it was trained and simulated systemically by using Matlab neuralnetwork toolbox. It was concluded that the model could be built by the BP neural network, cost-effectively, accurately and rapidly during far infrared drying of Agaricus bisporus within the trial stretch. It was found that the predictions of the artificial neural network model fit the experimental data preferably, and the applications of the artificial neural networks could be used for the online state estimation moisture content with more suitable and accuracy.