摘要
分类是一种重要的数据分析技术,可以用于提取描述重要数据类的模型和预测未来的数据趋势。提出了一种新的粗集理论和神经网络融合的分类方法。不同以往的组合方法,在该方法中粗集理论是挖掘分类知识的主体,最终的分类规则由粗集方法从约简后的决策表中得到,神经网络只是作为一种辅助工具,用来对决策表进行属性约简,并通过删除网络不能分类的数据来对决策表中的噪声进行过滤。与以往方法相比,该方法在保留神经网络高鲁棒性的同时避免了从神经网络中抽取规则的困难。
Sortation, which can be used to extract, describe major data type model and forecast the data tendency, is an important technology for data analysis. The article starts to propose a new sortable data-mining algorithm integrated with a new rough sets theorem and the neural network. Different from the previous grouping method, rough sets theorem is the main part in mining sortation knowledge. The ultimate sotatable principle can be obtained from the decision-making table which is reduced from rough sets. The neural network is only an assistant tool which can be used to reduce attribute of decision-making table. Furthermore, the tool can filter the table noise through deleting the insortable data. Compared to previous skill, this technology can remain the high stability of the neural network. At the same time, it can avoid the difficulty to extract the rule from the neural network.
出处
《三明学院学报》
2005年第2期185-189,共5页
Journal of Sanming University
