| In the ensemble learning,there are two problems that the classifiers cannot effectively process data with low quality and the weak classifier affects the overall ensemble effect.Some new ensemble learnings are constructed by integrating the three-way decisions into different processes of ensemble learning.One of these ideas is applied to intrusion detection.By optimizing the data set attribute selection method to improve the classification effect of base classifier,a three-way selection random forest algorithm based on decision boundary entropy is proposed.Based on random forest,a method of attribute importance measurement based on decision boundary entropy is proposed according to the characteristics of data set attributes.A three-way attribute random selection algorithm is proposed by integrating the three-way decisions to maintain the attributes’ randomness and reduce the redundant attributes’ influence on results.Experiments on UCI data sets show that the algorithm performs well on small and multi-classification data.Three-way selection random forest algorithm is applied to intrusion detection.The three-way random selection rules of attributes in intrusion detection data are established according to their own characteristics to increase the probability and randomness of important attributes.The decision trees are trained with the obtained data.The experimental effect is verified,and a new method for intrusion detection is provided.To enhance the ensemble effect of the base classifier,a three-way selection ensemble learning model is constructed for the selection process of classifiers.The random consistency is eliminated by replacing accuracy with pure accuracy,and the three-way decisions are constructed by combining pure accuracy and diversity as the evaluation function of the classifier.Continue to make decisions on classifiers based on accuracy,and the optimal integration is determined according to the diversity eventually.Applying the model to random forest,and the three-way selection ensemble learning algorithm based on random forest is proposed.Experiments on UCI data sets show that the model can improve the accuracy of prediction and reduce the storage space of trees.Figure 21;Table 19;Reference 60... |
