A New Method Of Evidence Fusion And Its Application

In the face of low-quality data,inaccurate data and incomplete data,evidence theory has an essential advantage in dealing with such data.With the rise of artificial intelligence,in practical problems,various models and algorithms have many connections,but also have their own application scope,advantages and disadvantages.How to improve the advantages of models and algorithms has become a very meaningful research.The work of this paper is as follows:First,the paper constructs a new distance function for characterizing evidence by modifying the distance of evidence source,and verifies its rationality through experimental analysis.Based on this distance,a new synthesis algorithm is proposed.The experiment shows that this method can eliminate some deficiencies of evidence theory,reduce evidence conflict,and improve Focus rate.Secondly,in the MYCIN reasoning model,since the MYCIN model requires that the evidences are independent from each other,there is a relationship between the evidences applied in practical problems,and they are not equal and independent.In this paper,the MYCIN model is improved,and the improved MYCIN model combined with some construction ideas in Chapter 3 is applied in the synthesis to overcome some of its shortcomings.Thirdly,we do some research on image recognition by combining the algorithm proposed in Chapter 3 with classification algorithm.Firstly,two important features of the image are introduced,and the calculation of the weight and distance of the features is proposed.In the experimental simulation,the single feature is verified to be de classified,and the value and shortage of its features are described from the effect.Then,combining with multiple features in use classifier classification,the final result is obtained by constructing the function,and finally the improved algorithm synthesis in Chapter 3.This process overcomes the instability of single feature recognition and improves the recognition accuracy step by step.