This paper is concerned with the study, optimization and control of the moisture sorption kinetics of agricultural products at temperatures typically found in processing and storage. A nonlinear autoregressive with exogenous inputs (NARX) neural network was developed to predict moisture sorption kinetics and consequently equilibrium moisture contents of shiitake mushrooms (Lentinula edodes (Berk.) Pegler) over a wide range of relative humidity and different temperatures. Sorption kinetic data of mushroom caps was separately generated using a continuous, gravimetric dynamic vapour sorption analyser at temperatures of 25-40 °C over a stepwise variation of relative humidity ranging from 0 to 85%. The predictive power of the neural network was based on physical data, namely relative humidity and temperature. The model was fed with a total of 4500 data points by dividing them into three subsets, namely, 70% of the data was used for training, 15% of the data for testing and 15% of the data for validation, randomly selected from the whole dataset. The NARX neural network was capable of precisely simulating equilibrium moisture contents of mushrooms derived from the dynamic vapour sorption kinetic data throughout the entire range of relative humidity.