The invention discloses a health state identification method based on exhaled gas marker detection in the technical field of health state identification methods, and the method comprises the steps: carrying out the communication with a measuring instrument, obtaining a measurement sample factor, carrying out the standardization processing of the measurement sample factor, combining historical data, performing training by adopting a depth adjustment algorithm of a random forest decision tree to obtain an optimal algorithm model of the characteristic factors and the disease degrees, judging whether the algorithm model training effect reaches a set threshold value or not by adopting the optimal algorithm model, if so, quickly predicting the measurement sample in real time, and if not, carrying out model training again, and outputting results after measurement sample prediction. According to the method, the health state of the human body is identified, the limitations of invasiveness, injury, inapplicability to early diagnosis and the like of a traditional chronic obstructive pulmonary disease diagnosis method are overcome, the method can be used for early screening of COPD, diseases can be found as soon as possible, the COPD death rate is effectively reduced, and the method has important application value and wide application prospects.