Research Of Temporal Association Rules Mining Based On Time Granularity

Data mining is one of the most popular research hotspots in the current environment,which needs data analysis to complete the analysis of information and data.Besides,the existing rules and decision-making contents will affect the final analysis results.Association rules are the most valuable research goals in data mining.In the early period,mining method of the association rule was relatively static.Moreover,it did not pay much attention to the time factor.But in the process of consolidation,many scholars find that the rules of digging out from real life are often influenced by time.It is concluded that association rules need to take account of time factors.But this finding never realized that time interval had effect on time data.Aiming at the fact that time is dynamic transformation,rules are relatively static,and the concept of time series association rules comes into being.Temporal association rule mining is based on the time granularity of the transaction data set based on year,month and day.Even though time factor has been included in the consideration category,it doesn't solve the problem of how to divide time granularity.Different time granularity will have a great impact on the number,quality and efficiency of association rule mining.At the present stage,the time granularity partition used in association rule mining is divided by human equal interval,and the attributes of data are not taken into account.Therefore,in order to improve the quality of mining,it is essential to study a dynamic granularity partition method with universal significance.On the basis of discussing the coherence theory of time granularity and so on,and in view of the present time granularity static partitioning method will make the same rule present different trends,the thesis considers that the thought of time granularity and the idea of clustering analysis are similar.Therefore,the first step of this thesis is to divide the time granularity by the method of clustering analysis,and then the fusion statistics are put forward.A time granularity dynamic partitioning method for analysis and self organizing map neural network is presented.Then this thesis adopts use time series association rule mining algorithm.On the premise of this research,according to the existing data of respiratory disease mortality and meteorological data in Wuwei,the correlation of the existing two kinds of data is analyzed,and the time granularity automatic division of time granularity is completed by the time granularity dynamic division method proposed in this thesis,and the time sequence association rule mining model is established.Meanwhile,in view of the hierarchical relationship between meteorological data,the concept of layered mining of experimental data is added to the algorithm of temporal association rules mining,which makes the algorithm more accurate extraction of valuable association rules.In this thesis,time series association rule algorithm was adopted to mine the experimental data,and the hidden correlation information between respiratory diseases and meteorological factors was found.The results indicate that the method of time granularity dynamic division of fusion statistical analysis and self-organizing mapping neural network is applied to mining the multi-level sequence association rule,which has the ability to show the change process of the rule more clearly and thoroughly.In order to improve the quality of time series association rules,the detailed change trend of rules in time can be mastered.