Meta-rule mining method for dynamic association rules based on wavelet transform
Fan Xu · Journal of Computer Applications · 2012
Concerning the problem that the forecast accuracy of the meta-rule mining in dynamic association rules is not high,this paper put forward a method for applying wavelet transform to meta-rule mining in dynamic association rules to improve the forecast accuracy of the rules.Firstly,the Daubechies wavelet was used to transform the support count of the dynamic meta-association rules.Secondly,the approximate part and detailed part could be extracted according to the multi-resolution characteristics of wavelet transform.And then the curve error calculation and selection control of wavelet decomposition level could be processed by using the two parts followed by inverse transforming and curve fitting by using the filtered approximate signal to conduct the predictions.The experimental results prove that the forecast precision is more than 90% by using the last forecast data.Finally,it turns out that the proposed method can better reflect the dynamic information and trends of the rules changing with time so as to get more accurate results of dynamic association rules with the guidance of reasonable meta-rules.