摘要
对多角度偏振成像仪测量偏差进行数据挖掘分析,解决多角度偏振成像仪测量偏差预测准确率低问题。利用包含决策树建立和决策树剪枝两个步骤的决策树算法,通过数据预处理,去除原始数据中的噪声,通过误差数据挖掘,选取真值方位和真值距离为输入属性,方位误差和距离误差为预测属性建立测量偏差模型,挖掘多角度偏振成像仪测量偏差;采用前向算法转换偏差测量结果为离散观察序列,概率较高序列便为测量偏差预警结果。通过实验分析,该方法在数据挖掘时具有较高的工作效率,与同类方法相比,该方法在预测多角度偏振成像仪测量偏差时准确度更高,平均绝对百分比误差更低。
In order to solve the low prediction accuracy of measurement deviation of multi angle polarization imager,data mining analysis is carried out for the measurement deviation of multi angle polarization imager.using decision tree algorithm including two steps of decision tree establishment and decision tree pruning,through data pre-processing,remove the noise in the original data,through error data mining,select true azimuth and true distance as the input attributes,azimuth error and distance error as the prediction attributes to establish measurement deviation model,mining the measurement deviation of multi angle polarization imager The forward algorithm is used to convert the deviation measurement results into discrete observation series,and the series with high probability are the measurement deviation early warning results.Through experimental analysis,this method has high efficiency in data mining.Compared with the similar methods,this method has higher accuracy and lower average absolute percentage error in predicting the measurement deviation of multi angle polarization imager.
作者
肖楠
刘斌
XIAO Nan;LIU Bin(Department of General Courses,Xi’an Traffic Engineering College,Xi’an 710300,China;School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,Xi’an 710021,China)
出处
《激光杂志》
CAS
北大核心
2021年第6期46-50,共5页
Laser Journal
基金
国家自然科学基金项目(No.NSFC61871260)。
关键词
多角度
偏振
成像仪
偏差
数据挖掘
决策树
multi-angle
the polarization
photography
deviation
data mining
the decision tree
