Research Of Mobile Application In Agriculture Information Service

Agriculture informationization, a major power to drive agriculture economy and technology forward, is also the trend of agriculture development of the world. As a result of global informationization, the phenomenon of "Knowledge Explosion" leads to the appearance of lots of internet web search engine. As a represent of agriculture web search engine, the website "Chinese Sounong" has been used widely and been rewarded. But at present, the informationization of countryside in China is still not developed enough, the lack of computer and network hardware result in that lots of farmer can not use these internet service in any time and any where. With the great development of mobile technology, the number of mobile phone user in China has exceeded 700 million. And the mobile phone is much more popular than computer and broadband hardware in countryside in China. As a result we could use mobile computing technology to make lots of farmers in China to fetch and publish agriculture information much more easily than ever before.As the great development of J2ME technology founded by SUN Inc, currently this technology has been supported by most of mobile device manufacturer from all of the world. For this reason, in this thesis we choose J2ME to design an agriculture information service platform. With this platform, people can query, read, publish relative agriculture information in any time and any where. Beside, we have a information-push-system built in this platform, this system can guess the potential interest of every user, and push the information that user needed to the mobile phone in time. In the server, we use SVM technology to classify the users interest.It is very important to determining the RBF kernel and error penalty parameters for support vector machines (SVMs). In this paper, we propose a method using the inter-cluster distances in the feature spaces and angle criterion that minimizes the angle between the kernel matrix and targets matrix to determine the optimum parameters for RBF kernel, then we use linear search method to determine the error penalty parameter. The final method we proposed is simple and easy to program. The experiment results show that the method costs much less computation time than other methods, thus the proper kernel parameters can be chosen much faster. And this method can choose proper kernel parameters with which the testing accuracy of trained SVMs is competitive to the standard ones.As an addition of mobile agriculture information platform, in this these we propose a inference engine that can be run on the J2ME platform. This inference engine can make up for the history that no inference engine can be found on the mobile platform. The knowledge base file that use in previous project of our group can also be used in this mobile inference engine. The experiment shows that this inference engine is light weight and very fast.