基于随机森林的亚健康状态预测与特征选择方法研究

摘要

Sub-health state is a low-quality status between healthiness and disease.The aim of the study is to determine which factors and/or combination of factors can predict sub-health state.In the paper,the authors carries out a clinical epidemiology survey and obtains 572 ca-ses (among them 523 are in sub-healthy state and the other 49 are in healthy state).There are 50 relevant symptoms included in report.They apply random forest categorization technique to predict the sub-health state based on clinical data analysis.They reach 9 1 .28%for the correct classification rates.The feature selections by 50-time random forest method (symptoms)are as follows:Fatigue,Palpitation,Myasthenia of limbs,Degree of fatigue,and Pessimism are important discriminative variables.Relevant experiments prove the practicability and efficiency of the proposed method.%亚健康状态是一种介于健康和疾病之间的低质量状态。研究的目的是要确定哪些因素或因素组合能够针对亚健康状态进行预测。临床流行病学调查，获取572个实际案例（其中，523宗均为亚健康状态，49宗为健康），在报告中包括了50种相关症状。应用随机森林分类技术进行基于临床数据分析的亚健康状态预测，正确分类率为91．28％。由50倍随机森林方法所得到的特征选择（症状），即疲劳、心悸、四肢肌无力、疲劳程度和悲观态度是重要的判别变量。相关实验结果显示了提出方法的可行性与高效性。