Since the complex industrial process data have characteristics of high dimension,nonlinear and noise,it is difficult to accurately classify and identify it.Therefore,this paper proposes an ensemble classifier(EC)method based on self-organizing feature map(SOM-EC).The useful process characteristic information(PCIs)can be obtained by using the self-organizing feature map network.Then,we use the PCIs as the input samples of EC.In this paper,the proposed method is applied to food safety early warning modeling process and oil-gas reservoir recognition process,respectively.The specific research contents are as follows:1.For the problem of data aliasing after projection transformation in the traditional Fisher discriminant analysis method,this paper proposes an improved Fisher discriminant analysis algorithm based on KNN algorithm(Fisher-KNN).By the judgment of boundary discriminant and using the KNN algorithm for the boundary data,the classification accuracy is improved.Moreover,the effectiveness of the proposed algorithm is verified by UCI data.2.For the complex industrial process data,this paper proposes an ensemble classifier method based on self-organizing feature map(SOM-EC).Firstly,the self-organizing feature map network is used to preprocess data,the centers of the cluster between attributes are extracted as the input sample of the ensemble classifier(EC).Then the EC consisting of SVM,MFBC,and Fisher-KNN is used to classify the input samples.Finally,the effectiveness and robustness of the proposed SOM-EC method is verified by UCI data set.3.The SOM-EC model proposed in this paper was applied to early warning modeling and analysis of food safety.Compared with the existing AHP-ELM method,the effectiveness of the proposed method is further proved.4.Using the SOM-EC model proposed in this paper,combined with the curve fitting(CF)method,we identified and analyzed the chromatogram data of heavy oil and light oil of an oil field in China. |