自适应条件下的相邻区分精度SVM集成模型研究

Study on SVM integration model of adjacent discrimination precision under adapti ve condition

摘要:

在数据进行集成的实际过程当中,分类器往往具有自主性,会随着样本数据的变化对自己进行 适 当调整,以此来提高自己的适应能力。对此,本研究以在数据样本区域内对相邻数据进行区分 的方法进行SVM集成方法研究,并最终提出了一种切实可行地支持SVM进行集成的方式。即针 对区分的数据样本区域,以一种新的搜索算法进行研究,利用FCM与模糊贴近度的结合来进行 计 算,实现在模糊特征空间集合频域自身位置的自动确定,再根据各项分类器的阈值数据系统自 行 录用当中的优异数据结果。并最终形成个体分器的数据结果从而进行集合性判定。结果显 示。在减少区分判断用时的前提下,这样一种数据算法能够达到提升分类器功能的有效作用 ；所建立的SVM集成模型具备动态自主适应性。集成过程当中分类器的个数选取关键点在于 分类精度阀值的选取,据此可以通过最优阀值的选取以达到模型判别能力的极大提升。

Abstract:

In the actual process of data integration,classifiers often have autonomy,and ad just themselves appropriately with the change of sample data,so as to improve th eir adaptability.In this paper,we study the method of SVM integration by disting uishing adjacent data in the data sample area,and finally propose a feasible way to support the integration of SVM.That is to say,a new search algorithm is used to study the differentiated data sample areas,and the combination of FCM and fu zzy proximity is used to calculate the location of the fuzzy feature space set a utomatically.Then,according to the threshold data of each classifier,let the sys tem employ the excellent data results by itself,and finally form the data result s of individual classifier,so as to make collective judgment.The results show th at such a data algorithm can improve the function of classifier effectively on t he premise of reducing the discriminant judgment time.And the established SVM in tegration model has dynamic autonomous adaptability.The key point of selecting t he number of classifiers in the integration process is the selection of the thre shold of classification accuracy,according to which the model discrimination abi lity can be greatly improved through the selection of the optimal threshold.