数字化集成信息增量学习算法仿真研究

Simulation Research on Incremental Learning Algorithm of Digital Integrated Information

传统集成信息增量学习算法无法更新数字化集成信息分类器,导致信息增量学习结果误差较大、效率偏低。为解决上述问题,提出基于深度学习的数字化集成信息增量学习算法。采用先验和条件两种概率获取后验概率后,获取样本类别标签。基于随机属性选择形成加权朴素贝叶斯分类器,分类数字化集成信息。并依据数据子集表更新该分类器。利用遗传算法获取更新后加权朴素贝叶斯分类器的最佳结果,完成数字化集成信息增量学习。实验测试结果表明,所提算法可有效控制训练样本数量,提升的信息增量学习效率,且该算法可较好地适应不同大小数据集。

Traditionally,the digital integration information classifier was unable to updated in common ways.Therefore,an incremental learning algorithm of digital integration information based on deep learning was proposed.In this research,priori probability and conditional probability were used to calculate the posterior probability,and thus to obtain the sample class label.Based on selection of stochastic attributes,a weighted naive Bayes classifier was designed to classify the digital integration information.Then the classifier was updated according to the data subset table.Furthermore,the genetic algorithm was adopted to obtain the best result of the weighted naive Bayes classifier after update.Finally,the incremental learning of digital integration information was completed.Experimental results show that the proposed algorithm can effectively control the number of training samples and improve the efficiency of incremental learning.Meanwhile,this algorithm can better adapt to different sizes of data sets.