Application And Research Of Hyper-Sphere Multi-class Support Vector Machine In Mutton Sheep Disease Diagnosis Expert System

In the mutton sheep disease diagnosis process, a veterinarian usuallyanalyzes the diease and obtains the corresponding diagnosis results accordingto the results of checking diseased sheep by medical devices, the obvioussymptoms, and their own pathological knowledge and experienceaccumulated over the years. Because there are various of mutton sheepdiseases and the less number of experts in remote areas, resulting in themutton sheep disease diagnosis often has deviation and delay in animalhusbandry farms. It is difficult to guarantee accuracy and extension ofapplication. In essence, the mutton sheep disease diagnosis expert system is apattern recognition and multi-classification problems on the basis of theexpert knowledge. According to the symptoms, mutton sheep diseases aredivided into13different kinds of disease categories. Different disease typesof samples are trained in multi-class learning machine and other samples arejudged by classmarks.Support vector machine (SVM) is a learning method based on VC dimension in statistical learning theory and structural risk minimization,which overcomes limitation of the diagnosis method based on traditional rules.It combines the best classification hyperplanes, quadratic programming, slackvariables and so on, overcoming the problems such as the dimension disaster,a local minimum and learning in traditional machine learning. SVM has agood performance in the classification. Because standard SVM is2-classifierand the actual problems mostly belong to multi-class problems, SVMpromotion to the multi-class becomes the research hot point in the field ofartificial intelligence and machine learning. At present, the main thought ismulti-class problems are devided into a series of2-class problems, and thenclassified by amount of SVMs. It needs to train amount of SVMs, so thelearning efficiency is relatively low and it does not suitable for multi-classclassification.In order to solve these problems, this paper adopts a direct type ofmulti-class classifier: hyper-sphere multi-class support vector machine(HSMC-SVM) to diagnosis mutton sheep diseases. It is great in capacity,rapid in training, strong in extension and directly classifies. HSMC-SVM establishes a super ball for each type of disease. Determining each supersphere is equivalent to find the solution of a quadratic programming problem.This paper establishes a HSMC-SVM sequential minimum optimizationalgorithm based on sequential minimum optimization thought in the trainingSVM and improves the heuristic choice method of the working set. The paperbuilds a mutton sheep disease diagnosis model to test. The result shows thatHSMC-SVM has advantages of timely, accurate and promotion efficiency inmutton sheep disease diagnosis.