Research On Data Mining System Building In Agricultural Information Grid

With the swift development of information technology and the unceasing accumulation of the data in agricultural area, the quantity of data becomes very huge, the uses feel more and more difficult to use these data. On the foundation of resources sharing in Agricultural data grid, using data mining technology for uses to extract the latent and useful knowledge becomes more and more important.This paper, bases on established agricultural data grid environment, by using existing data storage and computing capabilities, analyses and discusses the solution and tool in order to use data mining in the agricultural grid data, and then construct data mining system in agricultural data grid, aim to enhance the efficiency of agricultural scientific data, raise the database service level, and makes certain contribution for the agricultural data grid's further development.First, the paper analyses of agriculture information's characteristic: seasonal characteristic: localization, effectiveness, comprehensive, multi-level and so on., and discusses how to use data mining technology to solve the existing questions in several agricultural areas, like Meteorological, agricultural product market information, land management, output analysis chart, and irrigation. And then, based on detailed analysis of the data mining properties of the Agriculture Data Grid, builds an agriculture data mining system in agriculture data grid. The system consists of three main components: the Agriculture Data Mining Architecture (ADMA), the Agriculture Data Mining Toolkit (ADMT), and the Agriculture Data Mining Service (ADMS). ADMA describes the three-level architecture of data mining system; ADMT provides a large amount of data preprocessing and data mining algorithms; ADMS presents a data mining scheme to address the problems under grid environment through a form of grid service.In view of concrete algorithm, the paper at first analyses the characteristic of the rough sets algorithm, and according to the characters of this algorithm, make use of this algorithm into preprocessor. Then the author analyses two kinds of data mining algorithm: Clustering K-MEANS algorithm and Time series ARMA model, and then carries on the prototype experiment by using clustering and time series. Using k-means algorithm to solve the problem of Agricultural product sale decision-making, using ARMA model to solve the problem of forecasting the sale price of agricultural product. Finally, the paper summarizes the deficiency and existing problems in the research, and points out the perspective of further work.