Abstract
Text document clustering is to divide textual contents into clusters or groups. It received wide attention due to the vast amount of daily data from the Web. In the last decade, Meta-Heuristic (MH) techniques have been adopted to solve clustering problems. Motivated by that, the authors introduce a reliable version of the newly developed MH algorithm called Arithmetic Optimization Algorithm (AOA). Math arithmetic operators inspire the AOA: multiplication, subtraction adding, and division. The AOA showed good performance in several global problems; nonetheless, it suffers from entrapment in local optima in complicated and high dimensional problems. Therefore, this paper proposes an improved version of AOA for the text document clustering problem. The Improved AOA (IAOA) introduces an integration between Opposition-based learning (OBL) and Levy flight distribution (LFD) with AOA to tackle the limitations of the traditional AOA. The IAOA is examined with different UCI datasets for the text clustering problems and assessed with a set of CEC2019 benchmark functions as a global optimization algorithm with extensive comparison to existing optimization algorithms. Overall, experimental results show the superiority of the proposed IAOA compared to several optimization algorithms. Moreover, the proposed IAOA is compared with twenty-one state-of-the-art methods using thirty-one benchmark text datasets, and the results proved the superiority of the proposed IAOA.