A Feature Selection Framework for Anxiety Disorder Analysis Using a Novel Multiview Harris Hawk Optimization Algorithm

Machine learning (ML) has proved its success in medicine. One of the pending problems that ML can mitigate is anxiety mental disorder analysis. A better understanding of this disorder requires extensive analysis. The hurdle is that anxiety data is usually of high dimension, which frustrates the analysis task. Due to technological advances, medical data is gathered concerning different views, known as multiview data (MVD). Each view has its own data type and feature values, resulting in high heterogeneity. This article introduces a novel preprocessing feature selection (FS) algorithm, multiview harris hawk optimization (MHHO), to promisingly reduce the dimensionality of anxiety data which, in turn, eases the analysis task. The innovation of MHHO originates from utilizing a multiview linking strategy while combining it with the harris hawk optimization (HHO) algorithm's power. The HHO is used to locate the minimal optimal MVD feature subset, while multiview linking is utilized to figure out a promising fitness function to direct the HHO FS while taking care of all data views heterogeneity. MHHO is of OTHL2 complexity, where T is the iteration number, H is the number of involved harris hawks, and L is the number of objects. MHHO is heavily compared against seven recent competitors from its category for validation, using two publicly available anxiety MVDs. The experimental results show that MHHO is of high superiority in terms of convergence speed (converging in less than ten iterations), selected subset size (eliminating 75\% of the views; 66\% feature reduction) and high classification accuracy (approaching 100\%). Moreover, statistical comparisons show that MHHO is statistically different from its competitors, confirming its applicability. Finally, feature significance is calculated, shedding light on the most causing features of anxiety. The possibility of being infected by other diseases (depression and stress) is also investigated.