Balance-bagging-PRFS algorithm for feature optimization on insomnia data intervened by traditional Chinese Medicine

Abstract

Goal: Traditional Chinese Medicine (TCM) focuses on individual diagnosis. Besides the analysis methods on group level, clinical experimental data could also be researched with Information Technology to optimize the feature for individual healing effect; Method: we propose and apply a new method of feature optimization — Balance-Bagging-PRFS — to optimize the feature of insomnia intervened by TCM, aiming at solving problems typically in TCM data, such as mixing of discrete and continuous features and data imbalance; Result: from the view of all data, it is found that different levels of "ISI baseline score" and "Insomnia severity" have important influence on the curative effect. In treat group, different values of "environment" and "social field baseline" make remarkable difference on curative effect; while in control group, in which patients are treated with the placebo, "social field baseline", "survival quality baseline", and "classification of constitution" make sense; Conclusion: the method of Balance-Bagging-PRFS achieves good results in feature optimization for data from insomnia interfered by TCM, and it provides a basis for TCM individual diagnosis and for further optimization of symptom.