Research On Dynamic Association Rules Mining Based On Time Series

As a particularly important research method of data mining,dynamic association rules can describe their own characteristics over time.With regard to dynamic association rule,this paper has done the following major work.The paper is based on the research of correlative theories.Firstly,this paper combines the autoregressive integrated moving average model(ARIMA)of time series to model and analyze the meta-rules of dynamic association rules.The experimental results show that the ARIMA(2,2,1)model is more accurate in fitting the support countsequence and can fit the support count sequence.The relative error is controlled within 6%.Then,combining the GM(1,1)model,and the ARIMA-GM combination model is proposed,as well as the combination model of meta-rule support count is established.The experimental results show that the fitted curve of combination model can not only reflect the overall trend of the sequence,but also fully consider the detailed changes of the sequence,and ARIMA-GM combination model for comprehensive the advantages of each singel model show a higher precision.The relative error of the sequence is controlled within 4%.Finally,based on online sales data of a large supermarket in SQL Server database,this paper applies dynamic association rule algorithm to mine hidden association rules.According to the given threshold of support and confidence,frequent itemsets are found and strong association rules are mined.Then,the ARIMA-GM model proposed in this paper is used to model and analyze the support count of meta-rules.In conclusion,this paper focuses on the modeling of meta-rules of dynamic association rules.The ARIMA model,GM(1,1)model and ARIMA-GM model are used to model the support count of meta-rules,as well as the rationality and optimality of the model are verified.The results show that the proposed ARIMA-GM model can better reflect the trend of thesupport count sequence of meta-rules,and then help businessmen formulate reasonable and effective sales strategies.