Feature selection aims to select a small number of features from a large feature set to achieve similar or better classification performance than using all features. This paper develops a new binary particle swarm optimisation (PSO) algorithm (named PBPSO) based on which a new feature selection approach (PBPSOfs) is developed to reduce the number of features and increase the classification accuracy. The performance of PBPSOfs is compared with a standard binary PSO based feature selection algorithm (BPSOfs) and two traditional feature selection algorithms on 14 benchmark problems of varying difficulty. The results show that PBPSOfs can be successfully used for feature selection to select a small number of features and improve the classification performance over using all features. PBPSOfs further reduces the number of features selected by BPSOfs and simultaneously increases the classification accuracy, especially on datasets with a large number of features. Meanwhile, PBPSOfs achieves better performance than the two traditional feature selection algorithms. In addition, the results also show that PBPSO as a general binary optimisation technique can achieve better performance than standard binary PSO and uses less computational time.