High-dimensional data analysis and its great chances of overfitting result in great challenges for constructing efficient models in practical applications. To overcome these problems swarm intelligence algorithms can be utilized. However, the balance between global and local search throughout the course of a run is critical to the success of an intelligence optimization algorithm. Moreover, almost all the available algorithms are still having issues like premature convergence to local optimum and slow convergence rate, especially in high-dimensional space. As motivated above, a new hybrid optimization algorithm integrating particle swarm optimization(PSO) with gravitational search algorithm(GSA) is presented (denoted as PSOGSA). Based on the analysis of the compensatory advantages of the PSO and the GSA, in this paper, we integrate the ability of exploitation in PSO with the ability of exploration in GSA to update velocity equations. To update position equations a mobility factor is used which is guided by diversity of population to improve the final accuracy and the convergence speed of the PSOGSA. We also apply proposed algorithm to the cluster analysis of microarray data. Experiments are conducted on six benchmark test functions, four artificial data sets and three microarray data sets, and the results demonstrate that the proposed algorithm possess better robustness.