A Novel Particle Swarm Optimizer and Its Application to the Yield Curve Estimation Problem

Abstract

Particle swarm optimization (PSO) has been considered as one of the main swarm intelligence algorithms for solving single-objective optimization problem. How to update the velocities and positions of a swarm of particles is key to its optimization performance. In this paper, we propose to use the moving average of the local best positions visited so far as a key information to update the velocities and positions. Further, a central learning strategy is proposed in which the center of the local best positions is computed and used to update the global best position. Combining these two strategies with the updating formulas which are inspired by the free electron model in metal conductors placed in an external electric field, we name the proposed PSO algorithm as PSO with moving average and central learning strategies (dubbed as MAPSO). We test the performance of MAPSO on the CEC 2017 test problems with 10 and 30 dimensions. Experimental results show that MAPSO significantly outperforms some well-known PSOs in general. We then apply MAPSO on a very challenging real-world problem, i.e. the yield curve estimation problem, in macroeconomics. Our experimental study on the yield curve estimation problem with Shanghai interbank offered rates shows that MAPSO can effectively solve the problem and achieve the state-of-the-art performance.