Exploratory Partition Method of Continuous Attributes in Quantitative Association Analysis

Continuous attribute partitioning is the core problem and difficulty of numerical association analysis.For this reason,an iterative partitioning method of“first rough partitioning and association analysis,then users freely explore association rules and give partitioning suggestions,and then further partition according to the recommendations”is proposed.Based on the objective function that maximizes the confidence of adjacent regions and the subregion generation algorithm that satisfies the constraint of the region support threshold to provide candidate subregions with higher confidence.A visual analysis system is provided to allow users to further optimize the partition by combining existing rule data to select candidate sub-regions.Users can observe rule information and the relationship between rules based on scatter diagrams and chord diagrams to select the rules of interest;Further observe the detailed information of the selected rule in the histogram and select sub-regions to form a partition recommendation.Users can observe multiple rules and select candidate sub-regions respectively.In order to eliminate the differences and contradictions between the selected sub-regions,an interval integration algorithm based on the strategy of“divide first and then merge”and the three merging principles is proposed to form partition results and iterate.Through the use of a set of synthetic data sets,three sets of public data sets,and the Yunnan Province traffic violation accident data set for case study,high-confidence rules are obtained and the effectiveness of the proposed method is verified.