Research On High Utility Patterns Mining Based On Dynamic Indexed Lists

In recent years,the simplicity and efficiency of the utility list structure has led to the development of a number of algorithms based on this structure that have been used for a variety of tasks related to utility mining.One of the main limitations of high utility patterns mining based utility list algorithms was the high cost of concatenation operations between lists when the sets of items were merged.To address this problem,indexed list structures,windowed index structures and merged index structures were proposed in this paper.The information stored in the list can be accessed and updated quickly,the merging operation between itemsets was accelerated during the mining process based on the index values in the structure,and memory reuse strategies was employed to reduce memory consumption.The main contributions of this paper are as follows:(1)Research and implementation of the mining High Utility Patterns with Negative items in Incremental datasets(HUPNI)algorithm.When a new batch of data was incrementally inserted,an indexed list structure was applied to the algorithm and the new data for the item was quickly inserted into the previous structure without the original data being rescanned again,thus reducing runtime and memory.In addition,the utility values of the negative items were stored separately in the structure to ensure that the algorithm can mine High Utility Patterns(HUPs)containing negative items during execution.This algorithm addresses the limitation that traditional algorithms for mining HUPs with negative items can only handle static databases.(2)A sliding window-based algorithm for high utility patterns mining with negative items was investigated and implemented.The sliding index structure was used by the algorithm to store the utility and index value information of the itemset in each window,and when the window slides,the algorithm can quickly remove the old batch information and add new batch data at the same time.The algorithm solves the limitation that HUPNI can only handle batch data insertion.(3)Research and implementation of the incremental mining Closed High Utility Patterns with Negative items in Incremental datasets(CHUPNI)algorithm,which uses the merging of the index values of the original data and the new batches in the index structure to CHUPNI achieves early pruning and mines closed high utility patterns that were losslessly compressed.This algorithm solves the problem of large number of redundant patterns generated by the above two algorithms during the mining process.(4)The CHUPNI algorithm was used to mine HUPs in the shopping data of a supermarket dataset,and based on the results,advice was given to the shop on which items should be bundled or paired to improve the competitiveness of the shop;HUPs-based classification was studied and implemented to predict whether an item was on sale or not based on the closed HUPs mined in the product information.