〔Abstract〕 Objective To explore the feasibility of machine learning methods in the discrimination of tuberculosis patients. Methods The data of 15 observation indicators of 860 patients were obtained from a tertiary hospital. Through in-depth mining and analysis of the data, support vector machine, random forest and neural network model methods were used to discriminate the diseases of patients. Results The accuracies of the TB suspected patient discrimination models based on support vector machine, random forest and neural network were 90%, 91% and 88%, respectively. Conclusion All three machine learning methods can be used for discriminative analysis of suspected tuberculosis patients. In comparison, random forest performs better in discriminating patients with tuberculosis from those with pneumonia.