Improved PSO-SA Algorithm And Its Application In Association Rules Mining

Association rule mining algorithm is the most important part of the data mining Traditional mining algorithm is difficult to adapt to the current amount of data mining. People are trying to use the new intelligent algorithms, such as genetic algorithms, particle swarm optimization, fish swarm algorithm for mining association rules.Then this article combine the advantages of annealing algorithm and particle swarm algorithm, give a improved hybrid simulated annealing particle swarm algorithm. Pso algorithm is easy to fall into local optimal solution when it searchs the optimal solution,so it tend to work out the global optimal in actual problem. Because simulated annealing algorithm can accept bad solution at a certain probability, so it can be very good avoid algorithm trapped in local optimal solution.Algorithm implementation particles in updating its best position, the global best position and its position can have certain probability accepting bad solutions, so that can avoid particles trapped in local optimum.This paper applied the hybrid algorithm to the association rule mining, it use the Real-coded to set the particle,and select a appropria teevalution function. Finally, the mining association rules of the university fresh graduate employment situation for practical examples and genetic simulated annealing algorithm, the particle swarm algorithm compared with the hybrid algorithm in association rule mining viableand superiority. Finally, for mining association rules, after graduate in employment a number of recommendations:(1) Introduced the concept of data mining and association rule mining, and analyzes the advantages and disadvantages of the basic mining algorithm, described in detail the process of the realization of the particle swarm algorithm and simulated annealing algorithm and their advantages and disadvantages. The original group best location is recorded by two quantities, then when update each particle’s position and speed,the Metropolis criterion is used again.(2) Hybrid simulated annealing particle swarm algorithm’s implementation process is described in the paper. In basic particle swarm optimization algorithm, when update particle group of its own best location and best position, the Metropolis criterion is introduced. Metropolis is used twice can from two perspective of individual particles and the whole group to prevent algorithm trapped in local optimal solution, In this paper, the improved hybrid algorithm and the hybrid algorithm, the standard PSO algorithm of other literature are tested by several functions.(3) Improved hybrid simulated annealing particle swarm optimization algorithm is used in association rule mining. Graduate employment of association rule mining is regarded as practical example. Compared to GA and SA algorithm, the standard PSO algorithm in the same situation, under the condition of mining proved that the improved hybrid algorithm feasibility in association rules mining.