In the real world, the majority of complex optimization problems cannot be solved using classical methods. With the emergence of the internet and internet-based applications, a large amount of data is been generated on daily basis, which may contain a large number of features. Big data problems not only include a large number of instances and dimensions but the irrelevance, redundancy and noisy features are their main concern too. All the features associated with the data are not important. The presence of such features deteriorates the performance of any classification algorithm. So, it becomes important to select the feature subset to enhance classification in low as well as high dimensional data. Nature-inspired algorithms are found to be effective in finding the solution to such problems. In the present paper, a binary version of the chaotic-based hybrid nature-inspired algorithm for feature selection is proposed. Here, we select the best subset of features by hybridizing the Particle Swarm Optimization (PSO) algorithm and Harris Hawk’s algorithm (HHO) with chaotic tent map. The convergence and diversification of proposed algorithm is accelerated. To enhance the exploration of search space for avoiding local optima, we incorporated chaotic random variables generated through tent chaotic map function. By hybridizing PSO with HHO, the exploitation of algorithm is improved. The performance of the proposed method has been tested with eight well-known state-of-art nature-inspired algorithms over fifteen small as well as high-dimensional datasets on different evaluation measures. The analysis shows that the proposed algorithm proves to be better in comparison to all the competing algorithms.