Improved Fireworks Algorithm And Its Application In Near Infrared Spectra

Fireworks algorithm(FA)is a kind of Swarm Intelligence Algorithm(SIA),which can simulate the sparks' regular of fireworks blooming that exists in reality commonly.This algorithm not only has the ability of searching the high quality fireworks partially and carefully,but also can globally search on the poor quality fireworks.Significantly,the searching ability of FA is particularly outstanding of optimization issue and causes an extensive research.Although the traditional FA has an ability for Global Optimization,the massive number of explosions is still resulting in many problems,which include the slow convergence velocity and the low convergence precision.In this regard,scholars are carrying out the improvement on this algorithm.However,the depth of research is much lower than other SI Algorithm.Consequently,both the optimization and the application of FA have deeper significance.The main achievements of this paper are presented as follows:1.In order to improve the solution precision and the convergence speed of FA,based on the principle of FA,this paper tried to increase the population information communication.In addition,this paper combined the hill-climbing operator and collaborative operator with FA.It increases the diversity of populations by using the parallel computing of double populations.Moreover,this paper presented Double population Fireworks Algorithm(DFA).It uses population entropy to calculate the population diversity.It selected eight typical test functions to judge the optimization performance of the analysis algorithm.Experimental result proves that the fast convergence velocity,the high convergence precision,and the significant stability are obvious features of DFA.2.Directing against the demerits of FA,such as relapsing into local extremum,and particularly slow convergence velocity in the late evolutionary,this paper presents a Reabsorption based Fireworks Algorithm(RFA).In particularly,RFA attempts to introduce a reabsorption strategy into FA to expand the population diversity.it is based on the Reabsorption of Kidney Algorithm that can increase the idea of population diversity.In this chapter,eight typical test functions were used to analyze the algorithm performance and population diversity was calculated by the degree of particle dispersion simultaneously.Furthermore,T-test and Friedman Rank Sum Test were used for algorithm reliability evaluation.The experimental result indicates that RFA can solve the demerits of the late evolution of FA and has a higher precision.3.Owing to the constraints that exist on actual problem,this paper constrains the special equations.It reduces the variable dimension by parametric equations and transforms the constraint problem with an unconstrained optimization problem by quoting the penalty function method of Annealing Algorithm.FA can subsequently be used to solve the problem.Simulation experiment verifies that the new algorithm can solve the problem of constrained optimization,with strongly stability,fast convergence velocity,and high precision.4.In the field of Near Infrared(NIR),noise frequently associated with samples and equipment.It leads to a complex information on full spectrum and influences the quantitative model precision.Band Selection can extract the critical features of full spectrum,thus obtain a better quality model.After transforming this issue into an unconstrained optimization problem,using FA to build the quantitative model.The predict performance between full band and selected band can be comprehensively compared.There are four parameters are used for comparison,such as ratio of standard deviation of the validation set to standard error of prediction(PRD),Correlation Coefficient(RP),standard error of cross validation(SECV)and standard error of prediction(SEP).Experimental result describes that to compare with full band,FA can select feature band with better representativeness,higher performance and more accurate prediction.