Research On Classification Method Of Random Support Vector Machine And Its Application

Data classification has always been one of the hot research issues in data processing.Different algorithms have their own characteristics in data classification.The perceptron algorithm can effectively solve the linear separability problem.Through analyzing the duality problem in the paper,it is found that perceptron algorithm will influence the determination of hyperplane due to the different order of training samples and the initialization of parameters.The maximum interval classifier —support vector machine solves the problem.At the same time,in order to solve the linear inseparability problem and the noise problem,the kernel function and the slack variable are introduced,but a large number of hyperparameters are also brought.Although suitable hyperparameters can give support vector machines a high degree of generalization,support vector machines are highly sensitive to hyperparameters and require extensive debugging.Random forests do not need in-depth adjustment of parameters,and they have strong anti-noise ability and can also solve linear and nonlinear problems.However,random forests are affected by weak learners,and the performance of the algorithms is not very high.This paper first analyzes the advantages and disadvantages of random forests and support vector machines through standard data and visualization cases,and proposes a combination strategy of the two algorithms.It also proves the combination algorithm of random forest and support vector machine—a random support vector machine.Feasibility and advantages,and based on the first two algorithms reduce the complexity of the algorithm.In the end,the algorithm was applied to soil geochemical analysis data in the Milashan area of Tibet,and the data was first analyzed by random forest and support vector machine.In the classification of copper mines and non-copper mines,support vector machines have a good classification effect,but require a large number of adjustments.Random forest effects are not satisfactory,but it is easy to adjust parameters.The random support vector machine combines the advantages of the two algorithms and achieves an accuracy of 87.6% in the test set,which not only reduces the complexity of the algorithm,but also makes the classification effect better.