A New Model Of Information Fusion And Its Application

Information fusion is a comprehensive process of handling multi-information in multi-level. It is also a new subject which has universality and diversity. It has been used broadly not only in military affairs but also in civil domain.The main contributions in this paper include: 1)Realizing the parameter identification of the new model of information fusion; 2) Presenting a new method of case analysis and realizing its case-base maintenance(CBM) based on the above model . The details are as follows:Firstly, the new model of information fusion and its parameter identification are introduced in this paper. Choquet integral is a very popular information fusion approach. A major problem in application is how to determine a large number of fuzzy measure as the increasing number of attributes. In order to apply the above model, we realize its parameter identification by using the algorithm named Heuristic Least Mean Square(HLMS).Then we apply the proposed model of information fusion in a new method of case analysis and its CBM. The basic idea of CBR is that"similar problems have similar solutions". To represent and realize the generalization ability of CBR system, the key is the similarity measure among diverse cases. The frequently used model is to consider the weighted average of cases as the solution of the target problem.There are three crucial problems unsolved well when considering a CBR system as a map from the problem space to the solution space of the CBR. 1) How many solution-existing cases should be chosen for a new target problem? 2) The weighted average model can not deal with inherent interactions among diverse cases; 3) How can a designed CBR be efficiently performed for incomplete data when we must use it? We use the new model of information fusion to solve the above problems to some extent.