Uncertain Inference Based on Certainty Degree and Its Implementation by Artificial Neural Network
Minghui Shi · Jisuanji yingyong yanjiu · 2007
In this paper,we proposed an uncertain inference based on certainty degree(including certainty factor and certainty interval).A modified Back-Propagation(BP) Artificial Neural Network(ANN) is applied to realizing the uncertainty inference using two kinds of knowledge representation methods: certainty factor method and certainty interval method.First,the two methods of uncertainty knowledge representation are proposed.Next,BP network and its modified train function are pre￣sented.Then,how to apply the modified BP network to the uncertainty inference is illustrated in detail.Finally,an example for the certainty interval method using the MATLAB neural toolbox is given and analyzed.The simulation results show that ANN can play simple,but important and effective role in the uncertainty inference and that ANN can not only learn automatically experts' experiences,but also has the capability of generalizing the learned experience into more general situations according with experts' minds.