Research And Application Of Dynamic Search Fireworks Algorithm

The engineering design,scientific research and military management,and many other fields often involved in a lot of optimization problems.Moreover,in people's daily production and living practice,optimization problems is also almost everywhere,such as logistics,family financial management and so on.Generally speaking,optimization problem is to find the best solution to meet the specific requirements in many feasible programs.In order to solve these problems,scholars have studied a variety of optimization methods,such as Newton method,least square method,gradient descent method and Lagrange multiplier method and so on.However,with the development of production technology and the increase of human practical activities,many optimization problems encountered in real life are becoming more and more complicated,such as strong binding,multi-polarization.Early traditional optimization methods have been unable to meet the demand.Therefore,it is an urgent need to find a new optimization method with intelligent characteristics for complex optimization problems.In the mid-1950s,people get inspiration from the social behavior of animal groups and successfully put forward a number of stochastic optimization algorithms which is used to solve some complex optimization problems,such as particle swarm algorithm,genetic algorithm and bee colony algorithm.And these swarm intelligence optimization algorithms show great potential in solving complex optimization problems,and have strong adaptability and robustness,etc.Fireworks algorithm(FWA)is an emerging global optimization swarm intelligent algorithm developed in recent years.The algorithm was proposed by Professor Tan Ying et al in 2010,and the main idea of this algorithms the simulation of the natural phenomenon of the sparks produced by the fireworks explosion in the night sky.The fireworks algorithm has obvious advantages,such as fast convergence speed,easy to implement,and it also has the characteristics of explosive,diversity,simplicity and randomness.It has gradually attracted extensive attention of researchers at home and abroad and become a popular research direction in recent years.Fireworks algorithm is not only used to solve the optimization problem,but also has strong potential in practical engineering applications,It has been applied to image segmentation,the traveling salesman problem,0/1 knapsack problem,the optimization of distribution network reconfiguration,filter design and so on.But at present,the research and application of fireworks algorithm is very preliminary and is very superficial in some aspects,such as the interaction mechanism between fireworks and dynamic optimization problem solving.In addition,the fireworks algorithm still has some shortcomings,such as low precision and easy to fall into local optimum.Therefore,the improvement of fireworks algorithm and the expansion of application field has become a research focus.Firstly,the evolution of dynamic search fireworks algorithm(dynFWA),which is an improved algorithm of fireworks algorithm,is researched and analyze in this paper,and the algorithm is improved,then a dynamic search fireworks algorithm with learning factor and adaptive tendency is proposed,which is called improved dynamic search fireworks algorithm(IdynFWA).The learning factors in the improved algorithm make full use of the historical success information in the search process,so that the fireworks individual can learn from the "excellent" search information in the population,thus the size can be adjusted adaptively,and the two different learning factors can help to balance the local search and global search capabilities.In order to prove the optimization characteristics and effectiveness of the proposed algorithm,a set of standard test function optimization problems are carried out for the improved algorithm,the basic algorithm and other representative algorithms.The experimental results are compared and analyzed.The experimental results show that the improved dynamic search fireworks algorithm avoids the premature convergence of the basic algorithm to some extent,and the precision of searching optimization is also improved.Secondly,an improved dynamic search fireworks algorithm is applied to feature selection(also called feature extraction),and a feature selection method based on improved dynamic search fireworks algorithm is proposed.Feature selection refers to the process of selecting optimal feature subset from a complete set of original data features,and the specific evaluation criteria is used to evaluate the quality of the selected feature subset.Feature selection is widely used in many fields,such as machine learning,pattern recognition and data mining.It is also one of the key problems of the classifier classification.Each dimension of the final feasible solution generated by the method at the end of the subset generation phase is 0/1 integer coding,0 and 1 represent whether the corresponding feature is selected or not.Then,the classification error rate is used as fitness function to evaluate the quality of the selected feature subset.The fitness function value is also called the classification error rate is smaller,the better the selected subset of features is.Therefore,selecting the optimal feature subset is to find a feasible solution that minimizes the fitness function(classification error rate).A series of comparative experiments for the method and other feature selection methods are carried out in this paper.The results show that the feature selection method based on improved dynamic search fireworks algorithm can effectively remove the irrelevant feature and improve the classification accuracy rate.