排序方式: 共有120条查询结果,搜索用时 15 毫秒
51.
52.
53.
54.
The earliest resource dates back several decadeswhen Baumet al.[1]appliedthe algorithmin a specialcase of hidden Markov models.The most i mportantwork by Dempster,Laird and Rubin[2]formalizedex-pectation-maxi mization(EM)and proved its conver-gence.The basic EMalgorithmconsists of an expec-tation(E)step filling in missing data and finding itsdistribution giventhe known variables andthe currentparameter esti mates,and a maxi mization(M)stepre-esti mating the parameters to maxi mize the like… 相似文献
55.
针对Naive Bayes方法中条件独立性假设常常与实际相违背的情况,提出了CLIF_NB文本分类学习方法,利用互信息理论,计算特征属性之间的最大相关性概率,用变量集组合替代线性不可分属性,改善条件独立性假设的限制,并通过学习一系列分类器,缩小训练集中的分类错误,综合得出分类准确率较高的CLIF_NB分类器. 相似文献
56.
57.
数据挖掘在客户关系管理(CRM)中的应用 总被引:29,自引:1,他引:29
CRM现在逐渐成为企业的焦点 ,也成为计算机领域一个活跃的研究领域。介绍了有关CRM和数据挖掘的一些基本概念 ,以及数据挖掘应用于CRM所带来的好处 ;最后指出了如何在CRM中实施数据挖掘应用。 相似文献
58.
A new classification algorithm for web mining is proposed on the basis of general classification algorithm for data mining in order to implement personalized information services. The building tree method of detecting class threshold is used for construction of decision tree according to the concept of user expectation so as to find classification rules in different layers. Compared with the traditional C4.5 algorithm, the disadvantage of excessive adaptation in C4.5 has been improved so that classification results not only have much higher accuracy but also statistic meaning. 相似文献
60.
关联规则挖掘AprioriHybrid算法的研究和改进 总被引:5,自引:0,他引:5
分析了关联规则挖掘Apriori、AprioriTid和AprioriHybrid算法的优缺点,针对AprioriHybrid算法的瓶颈提出了一种使用支持度矩阵对频繁2项集快速验证的方法,并给出了一种简单易行,而又高效的逐步缩减交易数据库的方法,加快了对候选频繁k项集的验证速度,从而显著地提高AprioriHybrid算法的效率。 相似文献