Research On Feature Selection Method Based On Dragonfly Algorithm And Flower Pollination Algorithm

In the real world,our research is inseparable from a variety of large amounts of data.However,the imperfection of data acquisition technology and the existence of redundant noise in the collected data.If we want to improve the classification accuracy,reduce the complexity of the algorithm and reduce the storage space occupied in the operation process,we need to delete the irrelevant information in the collected data.Thus,we can achieve the effect of removing data noise.According to the existing common means,dimension reduction technology is often used in data denoising.By comparing the similarity between the original data set and the data after noise removal,dimension reduction technology can be divided into feature extraction and feature selection.Like feature selection,feature extraction and feature selection are based on the original features to find one or more features that can best distinguish the sample categories.With the development of computer technology and the progress of scientific research in China,the Internet and hardware equipment are developing rapidly in recent years,and a large number of data are naturally generated in daily life.The so-called era of big data is coming because these new data itself contains many valuable contents.People can help the existing fields of medical treatment,online shopping,finance and spatial information to further develop through in-depth study of these new data.In order to be able to use data in medical,online shopping,finance and spatial information fields,it is necessary tocollect and collate data.However,the data collected in reality contains a lot of redundant and irrelevant information,which is caused by equipment,technology and some non-human factors.These redundant and unrelated information will bring great difficulties to data representation,storage,analysis and processing,which may result in the inaccurate acquisition of the effective information contained in the data.Because of the large scale,high dimensionality and high redundancy of the collected data,data analysis technology in the fields of data mining and machine learning is facing great challenges.So,this paper adopts the feature selection method based on swarm intelligence optimization algorithm.Therefore,this paper mainly studies feature selection from the following four aspects:(1)Combining Dragonfly algorithm with flower pollination algorithm,after the dragonfly finds better food,further local development is made to improve the accuracy,and a dragonfly flower pollination algorithm is proposed.(2)In order to make the proposed algorithm applicable to feature selection,the fused algorithm is discretized.(3)A feature selection method based on the fusion of dragonfly algorithm and flower pollination algorithm is proposed,and a support vector machine(SVM)is used as a classifier,called discrete Dragonfly flower pollination algorithm.(4)The proposed method is applied to feature selection and compared with other similar methods.In order to verify the effectiveness and efficiency of the new feature selection algorithm proposed in this paper,six data sets commonly used in UCI datasets are selected to test the discrete Dragonfly flowerpollination algorithm.Finally,by comparing the classical and recent feature selection algorithms at home and abroad,we can find that the proposed algorithm is superior to the contrast algorithm in most data sets.Therefore,the proposed feature selection algorithm has good theoretical significance and application value.