Mining regular itemsets with interesting changes on regularity of occurrence

Nowadays, customers' behavior analysis is a crucial issue in competitive business. There is a need to know customers' behavior including changes of customers' behavior. This leads to an emergence of discovering itemsets based on their change on regularity of occurrence. However, this approach often generates a large amount of such itemsets causing users cannot directly utilize these itemsets to gain interesting information and/or knowledge. To address this issue, the task of mining regular itemsets with interesting changes on regularity of occurrence is introduced in this paper. A regularity and a change thresholds are thus assigned to define scope of interest and to filter uninteresting/insignificant itemsets based o n t heir change o n regularity of occurrence. To discover such itemsets, an efficient single-pass algorithm called Regular Itemsets with interesting Changes on Regularity of Occurrence Miner (RICROM) is presented. A new structure named New Interval Word Segment structure (NIWS) is designed to efficiently maintain occurrence information of each item/itemset during mining process. Experimental studies are conducted to show efficiency of R ICROM in the terms of runtime, memory usage and the number results in comparison with previous approaches.