Research On Automatic Modeling Method Of Optimal Structural Equation Based On Micro Search

Structural equation modeling,a research method based on statistical analysis technology,can simultaneously deal with the complex relationship between multiple variables and has been widely used in the field of social sciences.Users need to manually specify the model and use professional software for parameter estimation and fitting calculation when using structural equation modeling.The process is complicated and inefficient and requires users to have professional knowledge background.Current professional software in this field such as AMOS,LISREL,and MPLUS are developed by foreign companies,resulting in problems with copyright restrictions.Therefore,the automatic generation of structural equation modeling has significant research significance and application value.It is difficult to specify the structural equation modeling,which is due to the phenomenon that there are too many optimization indicators and mutual inhibition between indicators.To solve these difficulties,we model structural equation modeling specification as a multiobjective optimization problem,and then optimize multiple statistical fitting indicators of the structural equation modeling at the same time.The structural equation modeling specification in its nature is a search process.Besides,the size of the search space increases exponentially with the increase in the number of latent variables.The paper proposes an automatic modeling method of optimal structural equation based on micro search.Micro search is a method to solve the optimization problem by replacing the search in the entire decision set with the search in the small effective decision subset.The key is to construct a small effective decision subset.We observed that a set of better structural equation modeling has a public structure,and that the fitting index of the public structure meets the statistical requirements.The paper proposes the hypothesis that the public structure of the structural equation modeling has a major influence on the model specification.The combination of causal relations among variables in the public structure constitutes an effective decision subset for the search problem of the optimal structural equation modeling.We designed a method for calculating the weights of variable causality based on the non-independence between the latent variables when searching for the effective decision subset space.If changing the causality of a variable can improve the fit of the structural equation modeling,we will increase the causality weight of the variable,otherwise we will reduce the weight.The search algorithm first selects variable causality according to the weight,then constructs an effective decision subset by the permutation and combination of variable causality,and finally searches for public structures in the effective decision subset.We use the data in structural equation books for testing.Experimental results show that the automatic modeling method of optimal structural equation based on micro search can search for a structural equation modeling that meets the statistical requirements in the effective decision subset.Compared with the original search space with a size of 2^?2,the size of the effective decision subset is only 2?.Based on the above research,the paper designs a system for automatically generating structural equation modeling.Compared with existing professional software,this system automates the design of structural equation modeling.It allows users without professional domain knowledge to obtain the optimal structural equation modeling.It reduces professional requirements and labor costs.It also promotes the localization of software in this field.