摘要
针对时序动态关联规则挖掘中,支持度向量在时间特性上不宜观察其整体变化趋势与预测问题,提出了把小波变换应用到动态关联规则挖掘中,并建立自回归模型预测其整体趋势变化。该方法首先利用小波技术对挖掘出的动态关联规则支持度向量进行变换;其次,通过小波变换的多分辨率特点提取出近似部分和细节部分,同时利用近似部分观察其整体趋势变化;然后对两个部分分别进行重构,并利用自回归预测模型对重构的两个部分进行预测和累加,从而得到最后的预测结果,并用绝对误差检验其预测的精准度;最后实验证明其预测精度较高。
To solve the problem that the support vector was difficult to observe its overall change tendency and forecast on the time series in dynamic association rules mining,a method of wavelet transform applied to dynamic association rules mining was put forward,and auto-regression( AR) model was built to observe its overall change tendency and forecast. Firstly,the wavelet was used to transform the support vector of the dynamic association rules. Secondly,the approximate part and detail part were extracted respectively according to the multi-resolution characteristics of wavelet transform,and the chang trendency of the rules was reflected by approximate part. Then,the two parts were reconstructed respectively. The final forecast results were the sum of two parts which was respectively predicted by AR model. The absolute error was used to check its prediction accuracy. The prediction accuracy proved by the experiment is high.
出处
《河南科技大学学报(自然科学版)》
CAS
北大核心
2015年第6期40-45,7,共6页
Journal of Henan University of Science And Technology:Natural Science
基金
国家自然科学基金项目(61170024)
中央高校基本科研专项基金项目(121031)
关键词
小波变换
动态关联规则
自回归模型
wavelet transform
dynamic association rules
auto-regression(AR) model