Severe Convective Weather Identification Based On Data Mining Technology

As a highly destructive weather,severe convective weather often causes serious losses to our country's economy,agriculture and people's livelihood.Therefore,this paper proposes a severe convective weather identification method based on data mining technology by analyzing the reflectivity image and profile of Doppler weather radar,which can effectively identify the severe convective weather and help meteorological researchers to achieve related tracking and monitoring.The main work involved in this method is as follows:(1)Convert the reflectivity factor in radar-based data into reflectivity image and vertical profile,then extract attributes from a total of 1926 severe convective weather samples,and add some widely used attributes about weather identification to construct a 27-dimensional attribute database.(2)Based on rough set theory,a hybrid attribute reduction algorithm by combing fish swarm algorithm and shuffled frog leaping algorithm(FSA-SFLAAR)is proposed.The efficiency and accuracy of this algorithm are verified by a series of experiments and comparative analysis on the UCI datasets.(3)Convert the attribute database into a weather identification decision table,discretize it by supervised discretization method,and then use FSA-SFLAAR to realize attribute reduction,and a decision table with 7-dimensional condition attributes and one-dimensional decision attribute is obtained.The redundant information is effectively removed.(4)Generate a CART decision tree based on the reduced decision table and establish a severe convective weather identification model.Considering that in the actual application,new severe convective weather samples will be continuously obtained,so the naive Bayesian method is applied to the decision tree to realize the incremental learning of the identification model.The experimental results show that the proposed severe convective weather identification method with self-learning can accurately determine the severe convective monomer echo region,realize the weather type judgment,and maintain the efficient incremental learning ability.In summary,the abovementioned method can effectively identify the severe convective weather,realize the classification of hailstone and rainstorm,and meet the needs of actual applications.