动态朴素贝叶斯网络分类器的特征子集选择

FEATURE SUBSET SELECTION FOR DYNAMIC NAIVE BAYESIAN NETWORK CLASSIFIER

分类准确性是分类器最重要的性能指标,特征子集选择是提高分类器分类准确性的一种有效方法。现有的特征子集选择方法主要针对静态分类器,缺少动态分类器特征子集选择方面的研究。首先给出具有连续属性的动态朴素贝叶斯网络分类器和动态分类准确性评价标准,在此基础上建立动态朴素贝叶斯网络分类器的特征子集选择方法,并使用真实宏观经济时序数据进行实验与分析。

Classification accuracy is the most important performance indicator of classifiers.Feature subset selection is an effective method for improving the classification accuracy of classifiers.Existing methods of feature subset selection are mainly for static classifiers,while the research on dynamic classifier feature subset selection is rare.In this paper,the dynamic nave Bayesian network classifier with continuous attributes and the accuracy evaluation criterion for dynamic classification are presented first.A selection method of feature subset of dynamic nave Bayesian network classifier is developed based on this,while the actual macroeconomic time series data are used to carry out the experiments and analyses.