Abstract: The TSP problem is an NPhard problem， and the solution time increases geometrically with the scale of the problem. How to build an accurate solution in a short time is a challenge. Because the firework algorithm can quickly converge in the solution process and can jump out of the local optimal solution, based on the firework algorithm, the resource allocation for explosions is improved, and two operators are innovatively proposed: explosion operator that discards nodes and mutation operator that discards paths. Then, a Randomized Best Insertion Fireworks Algorithm (RBIFWA) is designed by 
using a fireworks selection strategy combining elite and roulette. RBIFWA are compared with some classical algorithms such as the basic firework algorithm, the chaos firework algorithm, the discrete bat algorithm, and the adaptive simulated annealing ant colony algorithm. The results show that the RBIFWA algorithm is significantly better than other algorithms in the number of iterations, and the accuracy of the algorithm is closer to the known optimal solution. It is proved that the RBIFWA algorithm has better performance and solution quality in solving the TSP problem.

Key words: fireworks algorithm, randomized best insertion, TSP problem, resource allocation