Study On Diversity Driven Adaptive Particle Swarm Optimization Algorithm And Its Applications

Particle swarm optimization(PSO)with its characteristics of simple and efficient to cause the attention of many researchers,and is widely used in all kinds of optimization problems.Due to the rapid loss of population diversity,PSO is easy to fall into localoptimal areas when solving complex optimization problems.Therefore,improving the convergence speed and convergence precision are always two important goals of studying PSO.Based on this,this dissertation proposes a diversity-driven adaptive particle swarm optimization algorithm and studies its application.The specific work is as follows:This dissertation put forward the diversity driven adaptive particle swarm algorithm(DAPSO)which combines the linear decreasing inertia weight and diversity drive speed two strategies,the former is helpful for accelerating convergence,which is helpful to keep population diversity and avoid premature convergence.After the combination of the two strategies,DAPSO better able to achieve the balance of the exploration and development In order to verify the performance of DAPSO,the experimental results of multiple tests on different types of benchmark functions show that DAPSO significantly improves the convergence rate of solving quality and algorithm reliability of PSO in most optimization problems,especially in highdimensional optimization problems,and its performance does not significantly decrease with the increase of dimensions Compared with other PSOs algorithms in this paper,DAPSO algorithm achieves a better balance between convergence speed and solving quality.In order to study the application potential of diversity-driven adaptive particle swarm optimization,two DAPSO-based application models are constructed in this dissertation.One is a B-spline curve energy smoothing model based on DAPSO,which can smooth out curvature changes under given tolerance constraints.Compared with other algorithms,DAPSO not only presents better smoothing effect,but also effectively controls the deviation from the original curve.The other is to build a BDAPSO feature selection model based on binary DAPSO.In 10 groups of UCI data sets,the BDAPSO feature selection model can greatly reduce the size of feature subset without significantly reducing the classification accuracy.In comparison,the BDAPSO model is superior to ACO,GA,PSO and other seven algorithms.When the evaluators of different classification algorithms are packaged,BDAPSO still has good optimization performance,and BDAPSO feature selection algorithm has certain universality.