Cloud Model Theory And Its Application

Cloud model is an uncertain transformation model,which is widely used to the intelligent control and data mining fields,especially in the field of multi-attribute decision making and evaluation.Most of these applications transform the precise quantitative values into the qualitative language values.The reverse cloud algorithm achieves this transformation.So it is very important to study the reverse cloud algorithms.Inverse cloud algorithm is based on certain algorithm steps to restore the three digital features of the cloud model:Expectation,Entropy and Super-entropy.The existed algorithm of backward cloud on the basis of the principle of moment estimation.Based on the analysis of the commonly used backward cloud algorithm and this paper proposes a new algorithm based on the idea of”the loss of information in the method of moment estimation and the higher precision of maximum likelihood estimation".The experimental analysis results show that this algorithm can get numerical characteristic values of cloud in high precision and good effect.Classic cloud parameter estimation methods are the moment estimation method and maximum likelihood estimation method.The two methods are both under the premise that we use the sample mean as the estimation of expectation,then getting the moment estimation and maximum likelihood estimation of entropy and hyper-entropy.In assumption that the expectation is not an unknown parameter,but it is a random variable and known the distribution.We can apply the bayes theory to get the expectation's posterior distribution and posterior estimation.And then through the posterior expectation estimation getting to the entropy and hyper-entropys' posterior moment estimation and posterior maximum likelihood estimation.According to the mean square error criterion,compared the methods to classic moment estimation,maximum likelihood estimation,Posterior moment estimation and posteri-or maximum likelihood estimation.Finally,we can get the conclusion that posterior maximum likelihood estimation is best than others.