An inverse method for measurement of thermal diffusivity has been developed. The inverse method was developoed based on finite difference method and an artifical neural network. The training data for the neural network was generated using a finite differnce of conduction heat transfer in agricultural products. The input to neural network was temperature distribusion of the agricultural products during haeting or cooling process. The output of the neural network was a thermal diffusivity. The developed inverse method has been applied to predict the thermal diffusivity of avocado, mango and seweet potato. The value of predicted thermal diffusivity was validated using a hot water treatment. The predicted temperature of finite difference using predicted thermal diffusivities were compared with the experimental tempurature distrubstion during hot water treatment. The results showed that the predicted tempuratures were closed to the experimental temperatures with error less than 5%.