摘要
由于在众多数据中得到目标信息的难度较高,为此深入研究关联数据的一致性规则,提出一种RBF神经网络下关联数据一致性挖掘算法。按照数据属性创建不同类型的属性记录库,累计求和全部属性相似度,消除重复数据,利用加权局部多项式算法平滑数据,剔除数据噪声;将内容相关的条件函数依赖作为一致性约束条件,通过模式融合与实例融合获得关联数据一致性规律;把预处理数据与一致性规律拟作输入值,引入RBF(径向基函数)神经网络,经过网络训练明确网络架构和参数,使用梯度下降策略调节训练参数,若输出误差低于准许误差,即可得到关联数据一致性挖掘结果。仿真结果表明,所提方法一致性挖掘准确率高、速率快,具有较高的应用价值。
In order to deeply study the consistency rules of association data,an algorithm for mining association data consistency based on RBF neural network was proposed.Firstly,different types of attribute record libraries were constructed according to data attributes,and then the cumulative sum of the similarity of all attributes was used to eliminate duplicate data.Secondly,the weighted local polynomial algorithm was used to smooth the data,thus eliminating data noise.Moreover,the conditional function dependency was taken as the consistency constraint,and then the consistency rule of associated data could be obtained through pattern fusion and instance fusion.After that,the preprocessed data and the consistency rule were regarded as the input values.Meanwhile,RBF(Radial Basis Function)neural network was introduced.After network training,the network architecture and parameters were determined.Finally,the gradient descent strategy was adopted to adjust the training parameters.If the output error was lower than the allowable error,the consistent mining result of associated data could be obtained.Simulation results prove that the proposed method has high accuracy and speed of consistency mining,as well as high application value.
作者
董琴
杨涛
DONG Qin;YANG Tao(School of Information Engineering Yancheng Institute of Technology,Yancheng Jiangsu 224000,China;School of Automation,Northwestern Polytechnical University,Xian Shaanxi 710072,China)
出处
《计算机仿真》
北大核心
2023年第7期457-461,共5页
Computer Simulation
基金
2020年国家自然科学基金(面上项目)(62076215)。
