Research On Fuzzy Multi-criteria Decision Making Method In Big Data Classification

Data mining is an important part of big data,and classification is one of the core technologies of data mining.With the progress of research in recent years,there are more and more innovative classification algorithms,which are widely used in big data processing.For some big data processing tasks,there are dozens of related classification algorithms for researchers to choose.How to choose the appropriate algorithm from these algorithms is an important problem in the field of data mining.In addition,it is difficult to test whether the classification algorithm is reliable and efficient.Based on the above reasons,this paper studies the selection of classification algorithms in a specific data environment.The mainly contents of this paper are as follows:(1)Introduce and analyze the key technologies of MCDM,fuzzy set,fuzzy multiple attribute decision making(FMADM),aggregation operator and classification algorithms.The MCDM method is improved to suit the environment of model evaluation.The basic classification algorithms such as K-nearest neighbor(KNN)algorithm,neural network algorithm,Bayesian network algorithm,C4.5 algorithm and CPAR algorithm are also studied.(2)The K-nearest neighbor algorithm is improved,and the K-nearest neighbor algorithm is evaluated by the algorithm evaluation model.This paper introduces the basic concept of fuzzy set and improves KNN algorithm.On the basis of traditional KNN classification algorithm,it introduces the theory of fuzzy set and combines the fuzzy C-means algorithm.The membership degree calculation is used instead of Euclidean distance calculation,and the fuzzy C-means clustering sample data is used to improve the efficiency of the fuzzy KNN classification.After that,an algorithm evaluation model is established to evaluate the KNN algorithm,the algorithm used for the contrast experiment are evaluated to verify the reliability of the algorithm and the evaluation model.(3)In order to extend the algorithm evaluation model to the big data classification algorithm,the validity of the algorithm evaluation model in big data classification algorithm is proved.This paper uses the MapReduce technology of Hadoop framework to parallelize the improved fuzzy K-nearest neighbor algorithm.Five alternativealgorithms are used to process the same two sets of data sets,and the running results are obtained.Operating results include TPR,TNR,precision rate and so on.(4)According to the data operation results of multiple classification algorithms,the algorithm evaluation and selection model are used to evaluate the data by using the Analytic Hierarchy Process(AHP),TOPSIS method and multi-attribute decision making method based on MSM operator.After the evaluation results of the three evaluation methods are obtained,all the evaluation results are summarized and re-evaluated to obtain more reliable schemes.