Research On Attribute Reduction Of Meteorological Observation Data Based On Genetic Algorithm

With the continuous development of informatization,meteorological observation data collection and meteorological elements are increasing.Because there is no clear purpose in the process of collecting meteorological data,the change of meteorological phenomena is often related only with the meteorological elements collected,so the attribute redundancy of the collected meteorological data is large.Redundant attributes not only reduce the efficiency of data mining,but also reduce the accuracy of data mining.Therefore,attribute reduction is very important for collecting meteorological data.This paper focuses on the characteristics of meteorological data and studies the algorithm for the reduction of meteorological data attributes based on genetic algorithms:(1)The basic knowledge of rough set theory is introduced firstly,and the knowledge classification,information system and attribute reduction in rough set theory are outlined.Then the necessity of attribute reduction and other optimization algorithms is analyzed.At the same time,the related knowledge of genetic algorithm is expounded,which will provide preparation for further research on attribute reduction of meteorological data.(2)In order to solve the characteristics of redundant data and strong data correlation,an interval valued information system is proposed.Combined with adaptive genetic algorithm,an Attribute Reduction Algorithm of Interval-valued Information System based on Genetic Algorithm(ARIGA)was proposed.As a typical time series data,meteorological observation data has a strong correlation within a certain time range.If the meteorological observation data is discretized into single-valued data,the correlation between data may be weakened,resulting in omission of some knowledge.The ARIGA avoids the missing of continuous data knowledge by introducing the interval value similarity,and satisfies the division of the equivalent class of the single value data and the interval value data at the same time.Finally,the performance of ARIGA on attribute reduction of meteorological data was verified by experiments.(3)In order to solve the problem of premature convergence and slow convergence of genetic algorithm,a Co-evolutionary Adaptive Attribute Reduction Algorithm based on Elite Strategy(CAARES)was proposed.Select the top M dissimilar individuals with the largest fitness value from the population to form an elite pool.Before crossover operations,select an elite individual from the elite pool to complete the crossover operation,and use the individual individuals to guide the population to rapidly evolve and increase the convergence rate.The CAARES improves population diversity by improving genetic operators and introducing random population.Experiments show that the CAARES improves the diversity of population evolution and ensures the convergence efficiency.