摘要
在模式挖掘应用于生产智能分析与辅助决策过程中,为提高模式特征选择的准确性和敏感度,设计生产案例样本结构,对特征实体进行逻辑转化;采用时域参数,以时序等粒度函数为基础融合并处理原始数据,降低数据维度,完成数据清洗;选定距离特征评估法选择敏感特征,应用径向基函数,引入距离评估因子反应类内聚集程度,描述特征敏感程度,完成时序化敏感特征的选择,达到深度挖掘数据潜在敏感特征,提高数据变化模式表达能力和鲁棒性的目的.最后结合油田聚驱生产施工作业过程,利用方法实现多元复合并发生产异常模式的特征选择过程.
In the process of pattern mining applied in the production of intelligent analysis and decision support,in order to improve the accuracy and sensitivity of the model of feature selection,design production sample structure,We make logic transformation for feature entity;Using time domain parameters and based on the time granularity function of integrating and processing the original data,reducing the dimension of data to complete data cleaning.Select sensitive features through the distance characteristics evaluation method,using Radial basis function(RBFS),via distance evaluation factors to reflect the inner aggregation degree,describe the characteristics of sensitivity and accomplish the timing-sensitive feature selection.To achieve the purpose that excavate potential law in the data deeply,improve the expressing ability and robustness on the data changed model.Finally,in combination with polymer flooding construction process,utilizing the method to accomplish multivariate composite of cocurrent production of abnormal patterns of feature selection process.
作者
卢志刚
李春生
胡亚楠
张可佳
LU Zhi-gang;LI Chun-sheng;HU Ya-nan;ZHANG Ke-jia(School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China)
出处
《数学的实践与认识》
北大核心
2019年第7期161-169,共9页
Mathematics in Practice and Theory
基金
黑龙江省教育厅科研专项创新基金(2017YDL-12)
黑龙江省教育厅科研专项创新基金(2017YDL-12)
黑龙江省自然科学基金面上项目(F2015020)
省教育科研规划重点课题(GJB1215013)
关键词
模式挖掘
时域参数
径向基
特征选择
案例样本
pattern-mining
time domain parameter
radial basis
feature selection
case sample
