Research And Realization Of Precipitation Forecast Based On Weighted KNN Algorithm

With the rapid development of economy and society, the influence of precipitation, especially ofshort time precipitation on urban traffic control, sewer system operation and people’s activities isgrowing. As a key and difficult point in weather forecast, enhancing the precision of short timeprecipitation has always been the focus of most attention. Through statistic analysis of a large numberof historical meteorological data, we design a forecast model which is a feasible way to improve theaccuracy of weather forecast. Firstly, a forecast equation is established to compute and find out theneighbor samples and compute their distance, in which similarity forecast method is applied tostatistic and analysis the data of area NCEP reanalysis and historical rainfall. Secondly, a forecastmodel is established to make short-term24hours precipitation prediction, which uses the weightedKNN algorithm to classify the samples. Finally, the forecast model is studied, the algorithm is givenand a lot of experiments are implemented. The result indicates that the algorithm is efficient toimprove the prediction accuracy.The main results of the work and conclusions of the paper are as follows：(1)In this paper, we research and analysis the conditions of formation, the classification and theinfluence factors of the precipitation, et al. The similarity forecast method is introduced to statistic andanalysis the historical data and the factor fields are recommended progressively by comparing theexperimental results, then, the similarity forecast equation is established to find out the neighborsamples and compute their distance.(2) According to the needs and characteristics of precipitation forecast, the basic principle ofKNN algorithm is studied and the advantages and disadvantages of existing KNN algorithm areanalyzed. Consequently, a novel EI weighted KNN algorithm is designed to establish the forecastmodel of precipitation. The forecast fitting rate and Ts score of EI weighted KNN algorithm is betterthan the other KNN algorithms’ obvious. Experimental results show that the model’s precise rate forthe fine or rain weather forecasting is high, and it has a certain capability for the level of rainfallprediction. At the same time, it can find out the similar weather situation in history samples, whichprovides a reference to judge the future weather trends for forecaster.