The CSI Application System Based On Structural Equation Model Determinate Algorithm

With the gradual transformation of business operation idea and quality management idea,pursuing customer satisfaction has become more important as ever before.In order to win in the drastic competition,companies should treat customer satisfaction index as the highest goal.The first problem needing to solve for customer satisfaction management is how to measure the customer satisfaction index.When doing that,to build the evaluation model of customer satisfaction index,considering the independent variable and dependent variable comprehensively,is one of the problems needing to research. Meanwhile,as customer satisfaction index is a Multi-index evaluation system,how to establish the weights of variable factors and generate customer satisfaction index using scientific and reasonable method,is also a problem need to research.By summarizing and reviewing different methods of the building of customer satisfaction model and establishing of the index weights,this paper proposed the principles to build customer satisfaction model.We focus the analysis of two modeling methods for structural equation:LISERL algorithm and PLS method.By compareing and analyzing we find the drawbacks of them;then this paper proposed the improved method,namely the deterministic algorithm of Structural equation model.By adding a constraint with modular length to the structural variables,we obtain the constraint least square solution with modular length parameters(MCLS) for observation equations,which is proved as the optimal initial value.We can use our MCLS as the initial value in PLS iterative progress to improve its convergence. Meanwhile,we change the modular lengths of observation variables,so that the path coefficients between the structure variable and its observation variables satisfy prescription conditions;then we solve the structural equation model.The metheod above is just deterministic algorithm of SEM.Lastly,it is extended to multilayer model.Also,multi-group SEM has been investigated and united by generalized linear model with convex constraint.We implement the improved structural equation modeling technology in our customer satisfaction index software.The system is designed initially for CSI measure of Shandong Electric Power Corporation.Adopting B/S model,it provides simple and intelligence operation,various functions,and accurate data.In order to exhume efficient,novel,potentially available,and easy understanding digital information,the system deal with large-scale data sample using scientific computational method.The main parts which is involved include project,model,index,questionnaire,sample selection,computing and ntelligental analyze of data.Finally,the kernel data is demonstrated to enterprise managers in a visualized form of picture,which can give managers some help in respect of analysis and decision-making.The innovations of this paper can be showed in several aspects as follows:(1)Find the PLS iteration initial value under the unit vector restraint,enhances the PLS convergence rate,explains the PLS iteration convergence,and improves the unbiasedness of constraint least square solution.(2) Propose determinate algorithm of SEM with the least square of errors and prescription conditions.Then,Construct multilayer and multi-group SEM,solve their computation problem.(3) Centering on the determinate algorithm of structural equation model,we implement the syetem of enterprise customer satisfaction index from China's realities. This paper analyses the functions and processes of software in detail.