[Comparative study on two methods of back propagation network test in TCM syndrome typing of depression].

ABSTRACT

OBJECTIVE: To compare the effect of the two methods of back propagation network (BPN) test on TCM syndrome typing of depression. METHODS: Test was carried out by two methods as following (1) Cross train-test method: 1731 patients with depression typed to 5 syndrome types were randomly divided into 2 groups, and they were trained and tested in turn; (2) Round-Robin method: Test was conducted in an altered cycle mode, that is, in a cycle, one out of the 1731 patients were selected to be tested, while the others were trained, the next cycle started when the test on the selected patient was finished and another one for test was selected. In this way, one cycle after the other, until all patients had been tested. RESULT: The total training sensitivity of the two methods was 97.9% and 98.2% respectively, and the total testing sensitivity was 72.7% and 74.2% respectively. CONCLUSION: (1) The five TCM syndrome types of depression could be well differentiated by BPN, which is valuable for TCM syndrome typing in certain extent; (2) The sensitivity of Round-Robin method is slightly higher than that of Cross train-test method, but in comparison between them no remarkable significance was shown.