某型飞机燃油消耗随机森林模型的统计诊断  

Statistical diagnosis for random forest model of aircraft fuel consumption data

作　　者：王淑玲[1] 谢凤[2] 朱倩倩[1] WANG Shuling;XIE Feng;ZHU Qianqian(Department of Fundamental, Air Force Logistics College, Xuzhou 221000, China;Department of Aviation POL and Materials, Air Force Logistics College, Xuzhou 221000, China)

机构地区：[1]空军勤务学院基础部,徐州221000 [2]空军勤务学院航空油料物资系,徐州221000

出　　处：《黑龙江大学自然科学学报》2018年第1期61-64,共4页Journal of Natural Science of Heilongjiang University

基　　金：原总后勤部军需物资油料部项目(CX212C036);空军勤务学院青年科研基金项目(KY2016D007B)

摘　　要：为确定燃油消耗数据中可能存在的异常点及强影响点,运用随机森林算法,对预处理后的某场站近三年燃料油消耗数据建模;对回归模型分别做残差分析和影响分析,不仅从残差图中观察出偏离既定模型很大的数据点,还仿照经典的统计诊断理论,定义诊断强影响点的统计量,可确定出对统计推断影响特别大的点;所得结论与逐步回归法一致。In order to determine the outliers and strong influence points in aircraft fuel consumption data,the actual data of fuel oil consumption of a station for nearly three years are pretreated and built regression model based on random forest algorithm,regression model is studied by residual analysis and influence analysis. Not only the data points deviated from the established model are observed from the residual plot,but also the statistics of the strong influential point are defined by the classical statistical diagnosis theory.The point of the influence of statistical inference can be determined. The conclusion is drawn and the results are consistent with the findings of stepwise regression method.

关 键 词：燃油消耗 随机森林 异常点 强影响点 

分 类 号：O212.5[理学—概率论与数理统计]