The Research On The Fuzzy Reasoning System Kernel Based On B-spline Basis Function Approximation

In many areas, such as Artificial Intelligence, Machine Learning, Pattern Recognition and data mining, the knowledge in the intelligent systems has not only the general principles of law, but also a large number of experiences of experts. So, the knowledge is random, unreliable and fuzzy. So we have to deal with the problem with uncertainty and get correct judgments of the proposition. Uncertainty reasoning is one of focus research domain in Artificial Intelligence. The uncertainty knowledge representation and the uncertain reasoning need to be studied using the knowledge in the intelligence system.In this thesis, the background, basic concepts, research fields and progress of the uncertainty reasoning are outlined. The thesis consists of several parts of work as follows:(1) The uncertainty of the evidence and the uncertainty propagation is studied in the uncertainty reasoning. The layered knowledge representation structure is used in the reasoning process. This reflects the expert analysis and problem-solving ideas, and based on this knowledge structure, the basic reasoning mechanism is explained.(2) Membership function is the cornerstone of fuzzy reasoning, its form is various. The application of the reasoning methods that only concern the specific membership functions has limitations, and is unfavorable for reasoning system designing and expanding.(3) This thesis approximates the membership functions based on the properties of B-spline function. The system using this method only has to solve the operation and design of the basis function. It also has good expansibility and practicability. Experiments verify that the approximation error can be controlled by adjusting feature points using the curvature of the membership function. So, the precision of the whole reasoning method can also be controlled.