Research On Pattern Mining Algorithm Based On Evolutionary Computation

The value and significance of data lie in its ability to guide practice and discovering the value of data is inseparable from data mining technology.As a basic and important research branch in the field of data mining,pattern mining has been widely used in many fields.With the increase of data and data dimension,traditional pattern mining methods usually suffers from both high runtime and memory requirements.As a general optimization method,evolutionary computation(EC)has shown excellent performance in many practical application problems,and its research has almost penetrated into various fields.In recent years,using evolutionary computing to solve the problem of pattern mining has become a hot topic.This paper studies a more efficient high-utility itemset mining(HUIM)method,and a more complete pattern mining problem model and solution method.The existing EC-based HUIM methods can only mine partially high-utility itemsets(HUIs)that meet the conditions in a limited time.It is often time-consuming to mine all the HUIs.This problem will become more prominent with the decrease of the minimum utility threshold and the increase of the number of HUIs.In order to improve the efficiency of finding HUIs,a HUIM algorithm based on an improved binary particle swarm optimization(HUIM-IBPSO)is proposed.In HUIM-IBPSO,several strategies are proposed to improve the efficiency of mining HUIs,including neighborhood exploration strategy,restart strategy,particle movement direction adjustment strategy and repair strategy.In addition,a fitness value hashing technique is introduced to reduce the number of fitness evaluation for repeated particles and reduce the overall time-consuming of the algorithm.In order to further improve the efficiency of mining HUIs,a HUIM algorithm based on an improved genetic algorithm(HUIM-IGA)is proposed.In HUIM-IGA,a population diversity maintenance strategy is designed to maintain the diversity of the population and reduce the miss of HUIs during evolution process.In addition,an elite strategy is employed to prevent the loss of high-quality solutions.Experiments on real-life datasets show that the proposed HUIMIBPSO and HUIM-IGA are outperformed to state-of-the-art EC-based methods in terms of convergence speed,the number of HUIs found,and the time-consuming.In order to improve the completeness of the pattern mining problem model,mining patterns that appear frequently and completely in the transaction datasets,and have high utility values,to meet the actual needs of users in some real-life application scenarios,a threeobjective problem model for high-quality pattern mining is proposed in this paper,where the objectives are support,occupancy,and utility.An improved multi-objective evolutionary algorithm for highly qualified pattern mining(MOEA-PM)is then proposed to get a set of tradeoff solutions.A new population initialization is suggested in the proposed algorithm,which is used to ensure that the population is effectively distributed in the feasible solution space.By taking the properties of the model into consideration,an auxiliary tool is proposed and used to accelerate the convergence of the algorithm.Experimental results on real-world datasets show that the proposed three-objective problem model with the MOEA-PM algorithm can discover patterns that are both frequently occurring and has a high utility in the transaction datasets,while at the same time being relatively complete.Compared with the state-of-the-art MOEAbased pattern mining algorithms,MOEA-PM has better performance in terms of efficiency,quality of results,and convergence speed.