• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IIE Annual Conference Exposition >A Transformation Approach to Reduce Multicollinearity
【24h】

A Transformation Approach to Reduce Multicollinearity

机译：减少多重型性的转化方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

In many industrial applications, observational data is collected and stored to later become the focus of a modeling exercise usually for the purpose of process control and/or optimization optimization. However, statistical modeling with this type of data can be challeng challenging due to the presence of multicollinearity, strong relationships between inputs, which can seriously affect the precision of the estimated regression coefficients. In some cases, Principal Component Regression is used. However the major limitation of thi this technique is interpretation of the resulting principal components variables which is often difficult. Another common alternative is Ridge regression. This method decreases multicollinearity but introduces bias in the parameters estimates. In this paper a an innovative approach to decrease multicollinearty based on inputs transformations is illustrated illustrated. The proper input transformations are found by an evolution evolution-based computing algorithm that generates analytical expressions of the response as a function of the input variables by combining basic functions, inputs, and numerical constants constants. The main advantage of the proposed approach is that . It produces a model that is in terms of the original variables with reduced multicollinearity and without introducing bias in the parameter estimates.
机译：在许多工业应用中，收集了观察数据并将其存储到以后成为建模锻炼的焦点，通常用于过程控制和/或优化优化的目的。然而，由于存在多型性，输入之间的强烈关系，具有这种类型数据的统计建模可能是挑战挑战，输入之间的强烈关系，这可以严重影响估计的回归系数的精度。在某些情况下，使用主成分回归。然而，该技术的主要限制是解释结果的主要成分变量，这通常是困难的。另一个常见的替代方案是脊回归。该方法减少多元性性，但在参数估计中引入偏差。本文示出了一种基于输入变换的多元图减少多元图的创新方法。通过基于演化的进化的计算算法找到了适当的输入变换，该计算算法通过组合基本功能，输入和数字常量常量来生成作为输入变量的函数的响应的分析表达式。拟议方法的主要优点是。它产生了一种模型，即在具有减少的多色性的原始变量方面，并且在参数估计中不引入偏差而不引入偏差。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号