Abstract: Abstract: To study the process of tuberculosis transmission, a combination of data analysis and mechanism modeling, established three models that can more accurately describe the process of tuberculosis transmission. Using the data provided by China National Defense Control Center for nearly six years of tuberculosis, improved model based on SIS approximate analytical solution, Back Propagation neural network, and Locally weighted linear regression modeling and analysis. By using the variance comparison method for the obtained prediction data, it can be known, the improved model of the SIS approximate analytical solution, the variance between the predicted and actual values of the locally weighted linear regression model are small, it can more accurately predict the process of tuberculosis transmission. Simulation results show, these three modeling methods can provide a scientific basis for the prevention and control of tuberculosis, and within the sample range used in this article, compared to BP neural network model, Locally weighted linear regression model predicts more accurate, but the SIS approximate analytical solution improves the model with less predictive value fluctuations, it can be stabilized within a certain range.