Research Of Algorithm For Mining High Utility Itemsets Based On Vertical Pattern

Since the concept of data mining was proposed,data mining technology has been drawn more and more attention and gradually mature.Nowadays,all kinds of data mining technologies are widely used in real life and unconsciously change people's lifestyle.Pattern mining is a fundamental part of data mining technologies.Recently,the traditional frequent pattern mining could not meets people's realistic needs.For example,people are no longer to mine the frequent itemsets,but those itemsets with high importance or profits.In order to realize the goal,high utility mining becomes one of hot topics in frequent pattern mining,which is an importance extension of the traditional frequent pattern.Compared with the traditional pattern mining,high utility mining is not only used to mining the frequent itemsets,but also meets the actual need of the users.So high utility mining has good applications and wide application areas.Based on the frequent pattern mining,we study and analyze the current algorithm of high utility mining,and improve the correlative algorithms.In addition,the sequential pattern mining algorithm is applied to the lossless image compression algorithm LZW.The following content is our major works:(1)An efficient high utility mining algorithm based on vertical patterns,named IHUI-miner.To improve the efficiency of high utility itemsets mining is one of promising topics in data mining,for the high utility pattern mining is complex.The HUP-miner algorithm is typical and based on vertical pattern,although it reduces that total number of utility lists,each itemset of partitioned utility list needs more memory spaces.In order to solve this problem,we propose the IHUI-miner algorithm to implement the relationship of itemsets on the one extension which is based on the HUI-miner and reduce the number of utility lists.We consider the relevant relationship between the itemsets in the 1-extension of a itemset,and updates the remaining utility in the utlity lists of remaining itemsets.Meanwhile,the algorithm expands the LA-Pruning strategy of HUP-mining and deletes the part of non-remaining itemsets to decrease the TWU of itemsets.The results of experiments with different standard data outperform in the time consumption and list number.(2)An image compression algorithm named LZW-MAL algorithm,based on the maximal contiguous sequence pattern and the page replacement algorithm.LZW algorithm is one of the commonly used lossless compression algorithms in image compression.The method of constructing the dictionary adaptively constructed by the LZW algorithm is local rather than global.This paper introduces the maximal contiguous sequence pattern to make full use of the unused dictionary value,for replacing the maximal contiguous sequence of the image;meanwhile,we use the page replacement algorithm to improve the dictionary of LZW algorithm.We proposed a new image compression algorithm named LZW-MAL algorithm.The results of the experiences show that the LZW-MAL algorithm is better than the LZW algorithm in the image compression.