With the statistical theory used, an algorithmic model of temperature compensation based on the Gaussian process regression is established for infrared methane sensors. The model parameters are studied to fit the data and to minimize the fitting error. In the Matlab software platform, the model is trained with the Bayesian algorithm and the numerical simulation of temperature compensation is performed. The simulation results show that the model has small error and high accuracy, and it can compensate the nonlinear fluctuation of sensor signal at different temperatures.