The Study On Knowledge Discovery Based On Uncertainty Theory And Machine Learning

With the development of computer,communication and internet technology, data is developed with the scale of exponent, which is beyond the ability of people. How to find the useful knowledge or interesting knowledge from these data is the urgent problem. Right after this knowledge discovery happens to be born. The extracted and refined new pattern is obtained by knowledge discovery from these dataset, which is based on agriculture knowledge in this paper. The study on knowledge discovery is to obtain new pattern of agriculture knowledge or to improve traditional agriculture knowledge, which is to provide more better service for agriculture production.Firstly some situations are summarized on knowledge discovery, such as its occurrence,its development and its methods in this paper, which is to solve the existing problem of agriculture production by different methods of knowledge discovery, which mainly are rough set theory and Dempster-Shafer theory based on uncertainty theory and neural networks and support vector machine based on machine learning.Many attributes are mentioned during plant disease diagnosis, which contain root,stem,leaf,flower and fruit. Symptoms are complicated and similar among different diseases, which are difficult to discern for the non-professional. The constructing of knowledge discovery system based on rough set for Suli disease and the design of classifier for Suli disease based on neural networks are put forward in this paper in order that the non-professional can diagnose disease by complicated and similar symptom.The traditional nutrition diagnosis such as DRIS,M-DRIS,DOP has advantage and disadvantage respectively. It is difficult to make decision for user during diagnosis by these methods. Dempster-Shafer theory is very efficient fusion theory based on uncertainty theory. The different experts knowledge is fused by Dempster composition formula. It is good way to make decision for user. This paper takes advantage of Dempster-Shafer theory and constructs a fusion model on nutrition diagnosis methods applied to have a fusion on the three diagnosis methods. The result shows it is right and feasible.Because of the low fitting precision of the present model, this paper proposes the fitting method based on support vector regress machine for crop water production function. Compared with the present model, the result shows the method based on support vector regress machine is better remarkably.Existing extensions of rough set model are analyzed in this paper under incomplete information system. This paper modifies limited tolerance model proposed by Wang Guoyin as far as the limitation on relaxation and restriction of condition constraint is concerned. Compared by example, the modified model based on limited tolerance model shows that it is more applicable,reasonable and effective when the universe of discourse is partitioned into tolerance classifications. Its agriculture application will be the next.