Research And Application Of Naive Bayesian Classification Algorithm In Rainfall Prediction

With the continuous progress and development of society,all walks of life have a close relationship with weather forecasting,especially for the accurate forecasting of some severe weather.At the same time,with the rapid development of science and technology and the continuous improvement of meteorological observation technology,massive amounts of meteorological data have been accumulated,Which contains a large number of meteorological rules,how to obtain valuable information from meteorological data has become an important research content.In this paper,the forecasting problem of classification of rainfall levels in meteorological data by Naive Bayesian Classification algorithm is studied,point out the effectiveness of discretization on Naive Bayesian algorithm and the importance of attribute weighting to Naive Bayesian algorithm,to improve the performance of Naive Bayesian algorithm,from the discretization of unbalanced weather data and the Naive Bayesian assumption attributes are mutually independent to study.By analyzing the sample meteorological data and selecting the rainfall forecasting factors,a Naive Bayesian Classification model is established to complete the forecasting of the rainfall levels.This paper mainly studies the following:For the imbalance of rainfall distribution and the discretization of continuous numerical attributes in meteorological data sets,an improved class-attribute correlation discretization algorithm is proposed.After analyzing the two-dimensional quantization matrix of classes and attributes,the algorithm fully considers the distribution of meteorological data in attributes,and at the same time combines different discrete standards,an improved discretization criterion CAID is proposed,and then a CAID algorithm is designed and implemented based on the criterion.By selecting the best breakpoint from the candidate set of breakpoints,divide the attribute domain into several intervals,a more reasonable discretization scheme is obtained with the least information loss,and the recognition rate of a few rainfall classes is improved.Then,a Naive Bayesian Classification model based on the CAID discretization algorithm is constructed in combination with the actual meteorological data to realize the classification and prediction of rainfall levels.By analyzing and comparing the experimental results,It is verified that the proposed algorithm can better solve the discretization of continuous numerical attributes in meteorological data.Compared with other algorithms,there is better performance in subsequent classification predictions.For the problem that the Naive Bayesian Classification algorithm assumes that the attributes are mutually independent,in order to improve the classification and prediction accuracy of Naive Bayes algorithm,a Naive Bayesian Classification algorithm based on attribute weighting is proposed.The algorithm uses mutual information and conditional mutual information to comprehensively determine the weight of each attribute,thereby weakening the mutually independent assumptions of attributes.Then,based on the actual meteorological data after the CAID algorithm is discretized,a Naive Bayesian rainfall class prediction model based on attribute weighting is constructed.The experimental comparison and analysis results show that the Naive Bayesian Classification algorithm based on attribute weighting has a better classification performance,and to a certain extent,the accuracy of the Naive Bayesian model for predicting the classification of rainfall levels is improved.