This study introduces reverse direction supported particle swarm optimization (RDS‐PSO) with an adaptive regulation procedure. It benefits from identifying the global worst and global best particles to increase the diversity of the PSO. The velocity update equation of the original PSO was changed according to this idea. To control the impacts of the global best and global worst particles on the velocity update equation, the alpha parameter was added to the velocity update equation. Moreover, a procedure for diversity regulation based on cosine amplitude or max–min methods was introduced. Alpha value was changed adaptively with respect to this diversity measure. Besides, RDS‐PSO was implemented with both linearly increasing and decreasing inertia weight (with 1,000 and 2,000 iterations) in order to survey the effects of these variations on RDS‐PSO performances. Six most commonly used benchmark functions and three medical classification problems were selected as experimental data sets. All experimental results showed that when the grain searching ability is not so small in the last generations, the algorithm performance continues to increase. Experimental proof of it was showed up especially in RDS‐PSO using the cosine amplitude approach. Because the best results among all the RDS‐PSO types for decreasing inertia weight modes were obtained with 2,000 maximal iterations rather than 1,000 ones.