城市居民出行次数变权预测方法
|
摘要
传统居民出行生成次数采用线性回归分析现状数据,建立居民特征变量与出行次数间的关系,但并未区分样本权重,预测结果受样本点波动影响较大。因此,论文考虑对不同的样本点施加不同权重,以区分不同样本点对预测结果的影响程度,并以最小二乘参数估计法为基础,采用Robust估计中的Welsch方法构造样本点权重值,通过迭代运算确定样本权重系数,进而建立样本权重变化的预测方法。研究表明,变权预测方法可应用于样本量大、变量众多,并难以准确识别样本有效性的情况。预测结果有效地避免了数据波动对预测结论的干扰,可更贴近居民出行次数的变化趋势。
Linear regression is a traditional method to predict the trip frequency of urban residents,for establishing the relationship between the characteristic variable of the residents and the number of their trips.But for the weights of different samples were not distinguished,the prediction result is greatly influenced by the undulating of the samples.Concerned with the different influence degrees caused by samples to the result,each sample should be distributed specific weight.In this paper,based on the least square estimation,Welsch method of robust estimation is used to construct the weight value,then,a kind of iterative computation is adopted to determine the weight coefficients.As a result,a method predicting the changed sample weight is established.Study indicates that the variable weight prediction can be applied to the case that the number of the samples is huge,the variables are numerous,and the effectiveness of the samples is difficult to distinguish.The prediction result successfully avoids the disruption of data's undulating and is more approximate to the resident trip trend.
Linear regression is a traditional method to predict the trip frequency of urban residents,for establishing the relationship between the characteristic variable of the residents and the number of their trips.But for the weights of different samples were not distinguished,the prediction result is greatly influenced by the undulating of the samples.Concerned with the different influence degrees caused by samples to the result,each sample should be distributed specific weight.In this paper,based on the least square estimation,Welsch method of robust estimation is used to construct the weight value,then,a kind of iterative computation is adopted to determine the weight coefficients.As a result,a method predicting the changed sample weight is established.Study indicates that the variable weight prediction can be applied to the case that the number of the samples is huge,the variables are numerous,and the effectiveness of the samples is difficult to distinguish.The prediction result successfully avoids the disruption of data's undulating and is more approximate to the resident trip trend.
引文
[1]Cruz Ortiz M,Sarabia Luis A,Herrero Ana.Robust regression techniques a useful alternative for the detection of outlier data in chemical analysis[J].Talanta,2006,3(70):499-512.
[2]Paolo Pennacchi.Robust estimate of excitations in mechanical systems using M-estimators-Theoreticalbackground and numerical applications[J].Journal of Sound and Vibration,2008,310:923-946.
[3]杨马陵,等.地震前兆数据的稳健回归与建模[J].西北地震学报,1999,12:399-408.
[4]周蒂,等.稳健统计学与地球化学数据的统计分析[J].地球科学,1991,16(3):273-279.
[5]HAMPLE F R,et al.Robust Statistics:the Approach Based on Influence Function[M].New York:John Wiley&Sons,1986:20-21.
[6]S Morgenthaler,RE Welsch,A Zenide.Theory and Applications of Recent Robust Methods[M].Birkhauser,2003:40-52。
[2]Paolo Pennacchi.Robust estimate of excitations in mechanical systems using M-estimators-Theoreticalbackground and numerical applications[J].Journal of Sound and Vibration,2008,310:923-946.
[3]杨马陵,等.地震前兆数据的稳健回归与建模[J].西北地震学报,1999,12:399-408.
[4]周蒂,等.稳健统计学与地球化学数据的统计分析[J].地球科学,1991,16(3):273-279.
[5]HAMPLE F R,et al.Robust Statistics:the Approach Based on Influence Function[M].New York:John Wiley&Sons,1986:20-21.
[6]S Morgenthaler,RE Welsch,A Zenide.Theory and Applications of Recent Robust Methods[M].Birkhauser,2003:40-52。
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心