Fireworks Algorithm And Its Application

Traditional optimization algorithm is powerless when facing with increasingly complicated and enlarging engineering optimization problem. Swarm intelligence algorithm was raised by simulating simple and instinctive action of natural living beings, with the characteristic of easy operation, distributed computing mechanism, strong robustness, good scalability and broad adaptability, which has guided a new direction for solving those problems.Fireworks algorithm searches the surrounding area by simulating the exploding process of fireworks in the night sky, which combines the advantages of simple mechanism, easy operation and superior searching ability. However, it also possesses the disadvantages of slow convergence, low accuracy and easy to get extreme values. Due to the short history of fireworks algorithm, which lacks theoretical depth and limited application field. Further study is needed for fireworks algorithm as it's application in some fields is still blank, such as the discrete setting and big data clustering analysis problem. The main study contents of this paper are as following step:Firstly, this paper detailed introduces some mature swarm intelligent optimization algorithm in terms of principle, operational process and the research status at home and abroad. Meanwhile, it summarize some improving methods and application fields of those algorithms. And the paper also briefly introduces some innovative intelligent algorithms.Secondly, aiming at optimizing the drawbacks of slow convergence, low accuracy and easy to get extreme value, the improved fireworks algorithm modifies the method of explosion to increasing the diversity of population, adds the function for fully using fireworks beyond the boundary, introduces information interaction operator to fasten the swiftness of individual interaction. Improved fireworks algorithm is tested on benchmark functions, and the test result reflects that the performance of improved fireworks algorithm is obviously better than primary fireworks algorithm and some other classic intelligent algorithms in convergence rate and accuracy.Thirdly, discrete fireworks algorithm is proposed to solve the 0-1 knapsack problem as one of combinatorical optimization problem. Instead of using traditional Sigmoid function to discretize, the discrete fireworks algorithm adopts the code of discrete integer and greedy strategy is introduced into the proposed algorithm as well. Three data sets are used to test the proposed algorithm, the result proves that it is outstanding in the aspect of convergence rate and accuracy when compared with DPSO and GA algorithm.Fourthly, improved fireworks algorithm is applied to solve UCI clustering, which is tested by 3 data sets. It can be found that the clustering effect is outperformance when compared with K-mediods, SOPSO and GSO algorithms.Fifth, a novel algorithm to solve PPI network cluster is proposed which based on the mechanism of fireworks explosion. According to the network characteristics, energy and topology potential are defined for every node. The algorithm takes those nodes whose topology potential is more than zero with the fireworks explosions point in the explosive radius which is defined based on the theory of topology potential as the same class. The novel algorithm can improve those evaluation index of F-score, Avg.F and Accuracy when compared with MCODE, MCL, CPCA and COACH algorithms tested on four kinds of data sets.