Mining High Utility Itemsets Using Evolutionary Algorithms

Mining high utility itemsets(HUI)is an interesting research problem in the field of data mining and knowledge discovery.When database is great and has very many different item,we meet NP-hard.Evolutionary computation is an efficient way and able to find the optimal solutions.Recently,bio-inspired computing has attracted considerable attention,leading to the development of new algorithms for mining HUIs,such as HUPEumu-GARM?HUIM-PSO and so on.Although these algorithms are more efficient than traditional algorithms,they need to scan the database many times.Besides these algorithms just discover less HUIs but not all.In order to solve the problems,the main contents of the research include the following aspects:1)Based on artificial bee colony,a new algorithm named HUIM-ABC was proposed to mine HUIs.A bitmap is used to transform the original database;binary vectors represent nectar sources,three types of bee and itemsets.The PBVC and DNSG strategy make the algorithm more efficient,PBVC strategy check whether itemset is valid and DNSG dynamically adjusts the unreasonable itemsets.2)A general framework of evolutionary algorithms named Bio-HUIF was proposed to mining high-utility itemsets.A bitmap is used to transform the original database;the result of xor between different binary vectors is defined difference or distances of itemsets,adjusting the gbest of populations was proposed.Under the framework,three new algorithms named by Bio-HUIF-GA,Bio-HUIF-PSO and Bio-HUIF-BA are proposed based on GA,PSO and BA,respectively.Extensive tests conducted on publicly available datasets show that the proposed algorithms outperform existing state-of-the-art algorithms in terms of efficiency,quality of results and convergence speed.3)Two algorithms are proposed,one based on the standard PSO algorithm and the other that follows a newly proposed bio-inspired HUI framework,repectively named by HAUI-PSO&HAUI-PSOD.Two algorithms are proposed,one based on the standard PSO algorithm and the other that follows a newly proposed bio-inspired HUI framework.Experimental results show that the former is more efficient,whereas the latter can discover more HAUIs in relatively few iterations.