Equine diseases auxiliary diagnosis expert system based on fuzzy rule promotion theory

Abstract: Many diagnosis expert systems for equine had been developed in the past, but there are no in-depth studies for the equine disease diagnosis expert system. And also most of the traditional expert systems are used by the conventional inference approach. The domain expert confidence factor (CF) for each rule of diseases is kept unchanged with its original value in conventional inference approach. So most of the traditional expert systems have a static knowledge base with static inference, and the decision power of these systems remains same through the life cycle of the system. In fact, the progress of equine knowledge requires modification of the knowledge base and rule base in the system. In this paper, we suggested a new approach for providing the intelligence in the system for diagnosis of the equine diseases. This experiment was conducted to develop a remote auxiliary equine diseases diagnosis expert system. By collecting and analyzing the experiences of diagnosis and treatment from experts on equine disease, the numerical expression of the equine diseases diagnosis knowledge was developed. The knowledge of equine diseases was represented with the method called object-attribute-value triples act (referred as O-A-V act) that combined with the generative formula. As such, it was easy to extract knowledge rules and these rules were used for inference mechanism. Using the confidence factor, multi-valued logic was used to represent the rules of confidence level. In this paper, we suggested a new inference method which was based on use of a fuzzy rule promotion theory. This approach can enhance the intelligence of the disease diagnosis system. If a rule was repeatedly used in corrective diagnostic results, it was then promoted to a higher confidence factor by the rule promotion factor (PCF), and the PCF was the original confidence factor in the next diagnosis session. In short, the dynamic PCF which was generated in the past dialogue was used instead of static CF in the final decision making process. The dynamically promoted rules were derived from those diagnosis sessions, which resulted in successful decisions. This enabled more efficient decision making in the future sessions. With this approach, it was not only decreasing the number of interactive between the system and the users, but also leading to acceptable diagnostic results. Based on the research of knowledge representation and inference mechanism, an auxiliary diagnostic expert system of equine diseases based on Microsoft.Net and SQL Server 2008 was designed and developed. It provided online help to equine farmers and extension workers in China. For the inference engine of system, we used the fuzzy rule promotion methodology that matched facts against conditions and determined which rules were applicable. Also, it automatically revised the confidence factor of each rule. The system performance test was conducted by 804 disease test cases. The successful and unsuccessful diagnosis consultation sessions were noted each time. By thirteen iteration tests, it showed that diagnostic accuracy of the system was closed to 92.28%. It proved that the method was a new way to enhance the diagnostic intelligence. The system met basic requirement of users. Suggestion for future improvement is needed to modify the rule promotion theory by minimizing the errors in the estimation of confidence factor and then estimation of the promoted confidence factor of the rule. By constantly maintaining and updating the knowledge base, and enriching the knowledge base, the system can definitely have a wide application in the countrywide.