Risk Prediction And Influencing Factors Analysis Of Cost Consulting Enterprises

Cost Consulting Enterprise is one of the main parts of the construction market and an important part of the intermediary service system of the construction market.Zhejiang Province,as a major construction province,has experienced frequent activities of cost consulting enterprises.With the development of the construction market and the gradual improvement of its own management system,this type of enterprises has experienced a period of rapid development.However,due to the increase in the number of enterprises,the competition pressure increases,the homogeneity is serious,causes the partial enterprise not to conform to the development tidal current,the management failure.However,the enterprise survival is usually considered as the final standard to judge the enterprise performance.In order to obtain the profit and the stable project source in the market,the premise is to be able to withstand the market screening,to be able to survive.In order to make the cost consulting enterprise develop healthily and stably in the market,and to provide the strong basis for the stakeholders and the government decision-makers,it is necessary to analyze the survival law of this type of enterprise and its influencing factors.In this paper,the survival analysis model in statistics is used to study the operation of cost consulting enterprises.Based on the data of 775 cost consulting enterprises in Zhejiang Province from 2003 to 2018,we established Cox Proportional Hazard Model,deep survival neural network model and random survival forest model,this paper studies the influence of enterprise form,business scope,geographical position,patent,credit grade and so on on the survival time,and predicts the survival probability.By comparing the prediction results of three models with consistency index,we find that the order of prediction ability of the three models is: Random Survival Forest >Deep Survival Neural Network > Cox Proportional Hazard Model.At the same time,according to the importance score and forward variable selectionmethod,13 important variables were obtained for the random survival forest model,the three variables of capital,qualification grade and self risk are not selected from Cox Proportional Hazard Model,so 13 variables from random survival forest model are brought into Cox proportional hazard model,it was found that the predictive power of Cox proportional hazard model was not improved,so there might be a nonlinear relationship between these three variables and the response variables(survival state and survival time).Therefore,for the data set of cost consulting enterprises in Zhejiang Province,this paper thinks that the random survival forest model is better,not only improves the forecasting ability of the model,but also automatically identifies the nonlinear relations among variables,and sifts out the important nonlinear variables.