增量KNN模型的修剪策略研究

Research on Pruning Strategy of Incremental KNN Model

KNN模型是k近-邻算法的一种改进版本,IKNNModel算法实现了基于KNNModel的增量学习.然而随着增量步数的增加,IKNNModel算法生成模型簇的数量也在不断地增加,从而导致模型过于复杂,也增大了预测的时间花销.提出一种新颖的模型簇修剪策略,在增量学习过程中通过有效合并和删除多余的模型簇,在保证精度的同时降低了模型簇的数量,从而缩短了算法的预测时间.在一些公共数据集上的实验结果验证了本方法的有效性.

KNN Model is an improved version of the k-nearest neighborhood method.IKNNModel is a KNNModel based incremental learning method.However,the number of representatives generated from IKNNModel increases with the increase of incremental steps.It makes the generated model become more and more complex and decreases its efficiency for classification.This paper proposed a pruning strategy for IKNNModel via merging and removing some representatives during the process of incremental learning.Experimental results carried out on some public datasets show its efficiency and effectiveness of the proposed method.