The Rating Evaluation Model Of Medical Industry Supplier Based On Random Forest

At present,the traditional market economy environment is undergoing a great evolution.Enterprises are now facing not only the usual competition between enterprises,but to change the enterprise in the supply chain resources for contention.And the random forest algorithm has become a new method of using scientific operation.It is primarily used to discover information that can be used in a large number of data that can be manipulated in reality.In this paper,we choose the algorithm of random forest as the basis of the experiment.First of all,the research background,meaning and method of the paper are briefly described.At present,mostcompanies do not have a set of scientific systems when they do supplier evaluations,and are evaluated by their own experience.This method has a strong subjectivity,and the establishment of a reasonable assessment system for enterprises to reduce costs and reduce risk has great benefits.But the establishment of a good system needs to select those representative indicators.Through the study of domestic and foreign literature,the selection of indicators should be able to strictly fit the purpose of the study;the construction of the index system to meet:perfection,rationality,ease of operation.In the third chapter,the random forest algorithm model is described in detail.Random forest is a kind of tree classification combiner,the sample data processing using bagging and random selection of the way.And in the use of Bagging method of sampling,there will be part of the data will not be drawn,this part of the data can be used to estimate the model of the generalization error.At the same time,it is proved by experiment that the generalization error of the stochastic forest model converges to a finite value when the number of trees reaches a certain value,so the number of trees in the forest can be determined by this principle.According to the purpose of this study,23 index systems are selected.Because the random forest model can calculate the importance of the index system,the final index system can be obtained by experiment.Then in the last chapter is based on the index system to establish a random forest model,and through the random forest to verify its noise has a very good immunity.This method of evaluating the supplier through the establishment of the model is worthy of further research.The combination of practice and theory proves that random forests have very good performance.However,there are some problems in this study that merit further consideration.If the amount of data chosen is not large,there may be a deviation,and there is no treatment for the outliers,and the index system is not performed at the time of screening detailed explanation.