| As the rapid development of database and computer network, massive meteorologicaldata is recorded and stored by the meteorological department in their daily scientificresearch and work. It is very meaningful to make use of data mining to analyze themassive meteorological data for improving meteorological forecast and meteorologicaldata utilization. The classic algorithms of data mining have not got a good result toprocess massive data in accuracy and efficiency. The cloud computing brings highprobability for his efficient distributed computing and storage.This paper, on the basis of analysis of the problem in meteorological data mining andstorage, proposes a solution for massive meteorological data mining and storage. In thispaper, a meteorological data center is studied and designed based on Hadoop and Hive.In addition, the meteorological data center can satisfy the query statistics need ofbusiness units for its clear advantage in processing of massive meteorological dataproved by a test in this paper.On the basis of the data center, a solution is proposed in this paper which uses NativeBayes based on Hadoop to mining massive meteorological data. The whole processincludes data discretization, data reduction and data classification. Furthermore, thepaper, on the basis of analysis of MapReduce, rough set theory and Bayesianclassification, realizes a parallel method for computing equivalence class of rough set toreduce data and a parallel Native Bayes algorithm. Finally, compared with the classicNative Bayes, the proposed solution can improve the efficiency of mining massivemeteorological data, and get a better accuracy in classification which can bedemonstrated by the daily meteorological data experiment.The solution proposed in this paper not only solved the problem in the storage andmining of massive meteorological data to some extent, but also cost lower and expandeasily, which can be applied in daily meteorological data analysis as a accessorialmeans. |
PDF Full Text Request