One of the most difficult challenges in machine learning is the data attribute selection process. The main disadvantages of the classical optimization algorithms based attribute selection are local optima stagnation and slow convergence speed. This makes bio¬-inspired optimization algorithm a reliable alternative to alleviate these drawbacks. Whale optimization algorithm (WOA) is a recent bio-inspired algorithm, which is competitive to other swarm based algorithms. In this paper, a modified WOA algorithm is proposed to enhance the basic WOA performance. Furthermore, a wrapper attribute selection algorithm is proposed by integrating information gain as a preprocessing initialization phase. Experimental results based on twenty mathematical optimization functions demonstrate the stability and effectiveness of the modified WOA when compared to the basic WOA and the other three well-known algorithms. In addition, experimental results on nine UCI datasets show the ability of the novel wrapper attribute selection algorithm in selecting the most informative attributes for classification tasks.