Reduction Methods of Attributes Based on Binary Particle Swam with Simulated Annealing

This paper proposed a binary particle swam optimization method based on simulated annealing. The simulated annealing was introduced when particles updated their position. The algorithm convergence was controlled by adjusting the speed of annealing. The particles would not easily jump out of the ¿expected¿ search area when the fall of temperature was slow enough, which improved the particles' local search capability and made the optimization algorithm more efficient. This algorithm was applied to the attribute reduction of casing damage prediction attributes were reduced from original 62 to 12. The complexity of aftermath processing was significantly reduced.