摘要
为解决由训练样本局限性导致因素分析法中测试样本错误识别和无法识别的问题,进一步提高因素库对样本信息的利用,采用样本培育的方法,对离散决策表的修炼培育作了具体表述:利用初始训练得到的规则集对新增样本进行测试,加入反馈机制,对训练集进行多重训练,直至初始测试准确率不变.研究结果表明:样本培育方法能及时地用新增训练数据改写推理规则,更好地实现样本信息的利用价值.
In order to deal with the misrecognition and unrecognition in factor analysis because of incomplete samples, and further improve the utilization of the sample information, the method of sample cultivation is adopted. The cultivation of the discrete decision table is described in detail： using the rule set obtained from the initial training to test the new sample, the feedback mechanism is added, multiple training on the training set until the accuracy of the same. Numerical experiment results show that this algorithm can use the new training data in time to change the inference rules and achieve the purpose of sample cultivation better.
作者
曾繁慧
郑莉
ZENG Fanhui ZHENG Li(College of Science, Liaoning Technical University, Fuxin 123000, China Institute of Intelligence Engineering and Mathematics, Liaoning Technical University, Fuxin 123000, China)
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2017年第3期320-323,共4页
Journal of Liaoning Technical University (Natural Science)
基金
国家自然科学基金项目(61350003
71371091)
辽宁省教育厅科学技术研究一般项目(L2014133)
关键词
因素空间
因素库
因素分析法
样本培育
反馈
样本
错误识别
无法识别
factor spaces
factorial databases
factorial analysis
sample cultivation
feedback
sample
misrecognition
unrecognition
