针对基于内容的文本分类问题,提出了一种Agent与朴素贝叶斯分类(Naive Bayes)相融合的反馈分类模型和算法(Agent-NB).朴素贝叶斯分类是一种简单而有效的文本分类方法,然而有限大小的训练样本集,一般不具备良好的数据完备性,难以一次性构造出高性能的分类模型.基于Agent-NB的反馈分类模型,可结合Agent的智能特性,通过反馈学习过程,动态调整相应参数,使朴素贝叶斯分类模型不断逼近其理想模型,从而提高分类器的性能.实验结果表明,提出的Agent-NB分类方法,分类效果明显增强,召回卒、准确率和F1值与朴素贝叶斯分类算法相比有一定提高.%Aim at the content-based document classification, a feedback classification model and algorithm was proposed, which combines the Agent and the Naive Bayes classification algorithm. Although the Bayesian algorithm was effective, the limited size of training sample set generally do not have good data completeness. It was difficult to construct a one-time classification model with high-performance. For the feedback classification algorithm based on Agent- NB, the classification performance was improved significantly by the intelligence of Agent, the feedback of classification information, and the dynamic adjustment of classification parameters. The experimental results showed that the algorithm had excellent classification ability, the recall rate, accuracy, and F1 values were satisfied.
展开▼
