顾客满意指数分析方法研究若干进展  

Some Advances in Research of Analysis Method for Customer Satisfaction Index

作　　者：丁倩[1] 刘天桢[1] 

机构地区：[1]武汉大学城市设计学院,湖北 武汉

出　　处：《统计学与应用》2020年第5期808-816,共9页Statistical and Application

摘　　要：顾客满意指数(CSI)模型属于社会心理学范畴,需要根据观测样本进行测度计量,但是由于模型属于不确定性方程组,已有的算法和公式存在一些问题需要解决:CSI模型的偏最小二乘(PLS)算法不一定收敛,或者收敛速度太慢;一些文献给出的CSI最后计算公式有的缺乏统计稳健性,有的不够全面;现实工作提出了多层顾客满意指数模型问题,需要给出算法。本文介绍我们所在课题组在这些方面的研究所取得的成果:找到了PLS最佳迭代初值,极大提高了收敛速度;进一步给出了基于最小二乘和配方回归的CSI模型的确定性算法;给出了稳健而全面的CSI最终计算公式;给出了多层满意指数模型算法。成果已经在一些单位使用推广。The Customer Satisfaction Index (CSI) model belongs to the category of social psychology and needs to be measured based on observation samples. However, because the model belongs to the uncertainty equation system, there are some problems with existing algorithms and formulas that need to be solved. Partial least square (PLS) algorithm of CSI model does not convergence certainly or its convergence rate is too slow. Some papers offer the last formulae for CSI but which is not robust or not comprehensive. In actual work, the multi-layer CSI model and multi-group CSI model are proposed, but there are not suitable algorithms for them so far. This article introduces the achievements of our research group in these aspects: finding the best iterative initial value, enhancing the convergence rate, further more giving a definite algorithm;giving a suitable and comprehensive formula of CSI, giving suitable algorithms for multi-layer CSI model and multi-group CSI model. Our results have been used in many enterprises.

关 键 词：社会心理学 顾客满意度 偏最小二乘 确定性算法 多层模型 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]