多分类器系统的泛组合规则研究与应用-计算机工程与应用.PDFVIP

48 2012 ，48 （17） Computer Engineering and Applications 计算机工程与应用 多分类器系统的泛组合规则研究与应用 1 2 贾澎涛 ，何华灿 1 2 JIA Pengtao , HE Huacan 1. 西安科技大学 计算机科学与技术学院，西安 710054 2. 西北工业大学 计算机学院，西安 710072 1. School of Computer Science and Technology, Xi ’an University of Science and Technology, Xi ’an 710054, China 2. School of Computer, Northwestern Polytechnical University, Xi ’an 710072, China JIA Pengtao, HE Huacan. Research and application of universal combination rule of multiple classifiers sys- tem. Computer Engineering and Applications, 2012, 48 （17）：48-52. Abstract ：The combination operators of multiple classifiers system are fixed combination operator with relatively poor serviceability. The idea of flexibility of universal logic theory is introduced in multiple classifiers system, and a universal combination rule based on universal combination operation model is proposed. Universal combination rule is suitable to multiple classifiers system with parallel structure. Then genetic algorithm is used to estimate pa- rameters of universal combination rule. The experimental results on time series datasets show that the classification performance of universal combination rule is better than that of fixed combination rules, which are product rule, mean rule, median rule, max rule, min rule and majority vote rule. Key words ：universal combination rule; multiple classifiers system; universal combination operation model; genetic algorithm 摘 要：现有的多分类器系统采用固定的组合算子，适用性较差。将泛逻辑的柔性化思想引入多分类器系统 中，应用泛组合运算模型建立了泛组合规则。泛组合规则采用遗传算法进行参数估计，对并行结构的多分类 器系统具有良好的适用性。在时间序列数据集上的分类实验结果表明，泛组合规则的分类性能优于乘积规 则、均值规则、中值规则、最大规则、最小规则、投票规则等固定组合规则。 关键词：泛组合规则；多分类器系统；泛组合运算模型；遗传算法 文章编号：1002-8331（2012 ）17-0048-05 文献标识码：A 中图分类号：TP391 1 引言 择最合适的分类器进行分类，从而减少识别的错误 [3]