Global QoS optimal Web services selection is a NP complete problem.In order to overcome slow convergence of existing schemes

an improved discrete particle swarm optimization

called Discrete Particle Swarm Optimization with Non-Uniform Mutation Algorithm (short for DPSONUMA)

is proposed.In DPSONUMA

we firstly redefine the particle position

velocity and update operations to make the algorithm more suitable for this problem.Then the best particle is introduced a mutation ability

and a mutation probability function is also designed.And weight factors of velocity will adaptively change according to the fitness values

which can improve convergence.In addition

a local fit first strategy is introduced

which lead to both quicker converge and better results.Experimental results show DPSONUMA costs less time but higher quality components are obtained for composite web services.