In this text the authors present a study on the approach of estimating the lost gas during the retrieval process of coal sample within wellbore. A new method is proposed to improve the accuracy of prediction as the traditional methods often give inaccurate predictions. To this end, the major sources of errors with those traditional methods are first analysed. Then the pertinent modifications are made in this proposed method to have these errors minimised. With the present method the following improvements are made. (1) Physically, the transport equation used here for gas flow in the core includes both advection and diffusion effects; in contrast, the traditional methods are fundamentally based on a linear diffusion equation. (2) Geometrically, the present method considers the the actual shape of the core sample (say, a cylinder), while the routine counterparts have to simplify the geometry of the core to be a one-dimensional spherical object. (3) Numerically, the present method employs the actual retrieval history to specify the boundary conditions of the core sample, and uses a neural network technique to best fit the measurements with the numerical solutions obtained. In contrast, the routine counterparts need to assume a constant or a linearly declining boundary condition such that an analytical solution (in a form of a series) can be obtained. When the relevant sorption isothermal curve for the coal is known, the proposed model contains three parameters to be determined, which are related to the effective permeability, the effective diffusion coefficient, and the initial gas content, respectively. These parameters are determined through best-match of the experimental data by the neural network simulations. An application example is presented in this study and it shows that the proposed method can significantly improve the accuracy of prediction for the lost gas volume or the initial gas content than its traditional counterparts.