灰多元线性回归及其在搜索引擎中的应用

经典多元线性回归分析模型不能实时跟踪响应变量的动态变化和在大量样本中因出现病态数据而影响拟合效果的问题,基于灰色系统时间序列的特性,提出了将灰系统思想与经典多元线性回归分析模型结合,形成一种全新的模型——灰多元线性回归分析模型。实验结果表明,该模型不仅能更加准确地给出响应变量的变化趋势,而且能过滤掉少量病态数据,从而避免了对拟合效果的影响。将新模型应用到网站的搜索引擎中,通过对网站访问流量及各个关键字搜索频率,预测该网站下一阶段的网站的访问流量,预测结果可为网站的管理者提供决策支持的理论依据。

Grey Multivariate Linear Regression and Application in Search Engine

The classical multivariate linear regression contains two shortages: it cannot run after response variables,the abnormal data which the multivariate linear regression produces affect the simulation.To cope with the problem,based on the time sequence characteristic of the gray system,a new model,the grey multivariate linear regression model,which combines the gray system and multivariate linear regression analysis model is presented.The result turns out to be that the new model can make up the shortages of the classical multivariate linear regression analysis model.First,it can exactly present the expected value of the response variable and the change trend of the response variable.At the same time,it can filter out the abnormal data which affect the simulation.So it is an effective method.At last,we apply the grey multivariate linear regression to the search engine of network station.Through retreating the accessing flux of the network station and the searching frequency of the different in some times,we forecast the accessing flux of the network station in the next time.The result of the forecast would provide the theory of DSS(Decision Support System) to network station's super administrator.