Multi-objective Association Rule Mining Based On Improved Whale Optimization Algorithm

The whale optimization algorithm is a recently proposed metaheuristic optimization algorithm with excellent search performance,however,when dealing with high-dimensional optimization problems,a single optimization algorithm cannot yet completely avoid the problem of falling into local resolutions.In this thesis,the whale optimization algorithm is improved with other swarm intelligence algorithms in a parallel mixture and applied to the association rule mining optimization solution problem.The main work is as follows:1.An improvement of the basic whale optimization algorithm is implemented.The method improves the local search capability of the algorithm by optimizing the linearly varying parameters in the whale optimization algorithm with a nonlinear strategy to change their trend during the iterative process.The Lévy flight strategy is also implemented to update the position of the individuals,which enriches the diversity of the population and expands the search range of the algorithm.The experiments demonstrate that the improved Whale Optimization algorithm has a greater improvement in the search ability in comparison with the basic Whale Optimization algorithm on the 0-1 backpack problem for solving the maximum value problem.2.A hybrid algorithm model based on parallel strategy is designed,which mixes the improved whale optimization algorithm with other swarm intelligence optimization algorithms in parallel.Experiments showed in solving the maximum value problem on the 0-1 backpack problem,the hybrid algorithm had more significant improvement in solving the optimal value and a more obvious decrease in running time as compared to the single algorithm.3.The application of hybrid whale optimization algorithm in multi-objective association rule mining is investigated.In order to directly mine the association rules,the individuals in the algorithm population are encoded binary first,and the Michigan method is employed to the individual to generate the rules directly and mine the rules.Not only the support degree and confidence degree are selected as the evaluation standard of association rules while mining association rules,but also the lifting degree and certainty factor are additionally added as the multi-objective of optimization.The experiments prove that the hybrid algorithm can improve the mining efficiency and reduce the mining time when performing association rule mining over the basic whale optimization algorithm.In brief,this thesis focuses on the improvement of the basic whale optimization algorithm and mixes it with other swarm intelligence algorithms in the application of association rule mining.The relative experimental results of improved whale optimization algorithm mixed with swarm intelligence algorithm prove well performance in association rule mining problem,and whale optimization algorithm has a certain application prospect in this direction.