Improvement And Application Of Ensemble Learning Method Based On Support Vector Machin

Classification problems are frequently encountered in various fields of life,and solving such problems is rich in practical significance and full of challenges.Support vector machines and ensemble learning algorithms have been widely studied and applied in this area.Considering that these methods are not well studied in terms of how to precisely control the classification accuracy of the minority class when faced with the classification problem of imbalanced data.This paper improves the ensemble learning algorithm based on support vector machines.A novel balancing strategy is proposed to rebalance the weights during the boosting algorithm using the rebalanced sample weights to influence the training of the base classifier.This weighting strategy increases the sample weight of the misclassified minority while decreasing the sample weight of the misclassified majority until their distributions are even in each round.A series of base classifiers with different parameters will be generated,and the final classification model will be formed by a weighted combination.It is shown theoretically that the rebalancing strategy does not change the basic properties of the original standard Ada Boost algorithm model,it can continuously reduce the training error of classification on the training data set,and the training error also decreases at an exponential rate for the dichotomous classification problem.In addition,this paper introduces a new metric P-Mean as one of the evaluation criteria,which can better reflect the classification effectiveness of the model for the minority class.To illustrate the effectiveness and merits of the improved model in this paper,experiments were conducted to compare the proposed and comparison 10 models on 18 datasets in terms of six different metrics.Through comprehensive experimental findings,a statistical study is performed to verify the efficacy and usability of the proposed model.