Study On Diagnostic System Of Dog Disease Based On Intelligent Computing

This paper analyses the current states of research on medical treatment expert systems, especially those of animal disease diagnosis domain. To resolve the problems in the field, we did farther user's requirement analysis and studied the old disease diagnosis expert systems further, and some documents which correlative with them are lucubrated. On the basis of the above works, we put forward the idea and solution, i.e. to build dog disease diagnosis system based on intelligence computing.On account of knowledge acquisition, efficiently reasoning and self-study ability are the difficulty of diagnosis expert system development, and also they are the shortage of traditional disease diagnosis system. According to those shortage and the features of user's requirement and the function required by the dog disease intelligence system, the knowledge acquisition and reasoning mechanism of intelligence system is studied and designed, which uses rough sets theory, fuzzy theory and neural network as the key technologies. It can be described as follows.Knowledge acquisition . Use the attributes reduction algorithm based on rough sets to reduce the redundance attributes in the primary data of dog disease cases, which comes from domain experts and the cases in the data are vast and had been diagnosed. At the same time, we can get typical cases directly from dog disease diagnosis manual. Take the attributes, which are gained by above methods, as the input arguments to build neural network, which is based on rules and make the rules encoded explicitly. Using this neural network, reasoning processes can be understood and explained distinctly. Last, use the corresponding data of the cases training the neural network to acquire the diagnostic knowledge of dog disease and then create the knowledge base.Reasoning and explanation. When user select symptom, The corresponding network will be activation. Here, hybrid Inference Mechanism is used. The first inference result is arrayed by the value of posterior probability. If the conclusion is diagnosed, then output the conclusion. However, if it isn't diagnosed that means suspect conclusion, then continue the second diagnose. Another relational symptomsshow and can be choose by user, and then continue the second inference diagnose.Self-study. Disease clinical diagnoses which after the laboratory, may carry on the training, and perfect the knowledge library. If the system diagnosis is consistent with the clinical diagnosis, then optimized weight;Otherwise, it is inconsistent, then self-study and correct weight, make the knowledge library automatic consummation.Upon the foundation above, a dog disease intelligence diagnosis system based on rough sets theory, fuzzy theory and neural network, etc. is designed and realized with using many technologies, such as Java, Oracle Database, etc.