Analysis Of Meteorological Data Based On Data Mining

With the application of Chinese Meteorology, the meteorological fields haveaccumulated a large amount of data. There is a lot of important information behind theproliferation of data, how to take advantage of these data and find useful, but oftenoverlooked, has become an important task for researchers. The research and analysis ofthe meteorological data for the production practices and social life has become moreand more important. This paper processes and analysis the country meteorological dataof the northern Shaanxi by the decision tree method and time series analysis.Firstly, the temperature was assessed by the CART algorithm of the decision treemethod, where the main consideration month average pressure, average monthly vaporpressure, month minimum relative humidity, evaporation and rainfall were took intoaccount in the assessment. The decision tree model by the monthly average temperaturewas established; meanwhile, the decision tree model by the seasonal temperature wasestablished. By contrasted with the other research results, there was higher accuracy rateof the models in the temperature forecasting. The temperature discretization predictionmodel was established, for the specific values of the temperatures weren’t asked by theoutdoor operations department, which was used to predict the temperature. Bycomparing and studying this model with the previous two models, the accuracy rate wasimproved by this model in the temperature forecasting.Secondly, the simple seasonal models and seasonal ARIMA model in the timeseries were used to analyze and deal the meteorological data of northern Shaanxi countyground-based measurements, and the two models were compared and analyzed. Theresult was that the seasonal ARIMA model was better than the simple seasonal model ineffect.At last, the comparative analysis of the decision tree method with the time series prediction method was made in the air temperature applications. The result was that thedecision tree model has the lower prediction accuracy rate than ARIMA model in thewhole aspect, but it could be good to capture data mutation. And there were moreadvantages in the time series method when it was used to predict the continuous andsteady data.