A feature selection framework for anxiety disorder analysis using a novel multiview harris hawk optimization algorithm | JoVE Visualize

Area of Science:

Computational psychiatry
Machine learning in healthcare
Data science for mental health

Background:

Machine learning (ML) offers powerful tools for analyzing complex clinical data, aiding in diagnosis, treatment, and outcome prediction.
Anxiety disorder analysis presents a significant challenge, complicated by the multidimensional and heterogeneous nature of patient data.
The increasing availability of multiview data (MVD) in healthcare necessitates advanced methods for effective processing and analysis.

Purpose of the Study:

To introduce a novel preprocessing feature selection (FS) approach called multiview harris hawk optimization (MHHO) for anxiety data.
To reduce the dimensionality of anxiety MVD, thereby decreasing analytical complexity and improving processing efficiency.
To enhance the accuracy and reliability of ML models in analyzing anxiety disorders by effectively handling data heterogeneity.

Main Methods:

Developed MHHO by integrating a multiview linking methodology with the Harris Hawk Optimization (HHO) algorithm.
Utilized multiview linking to create a fitness function that accounts for the heterogeneity across different data views.
Employed HHO to identify the optimal, low-dimensional feature subset from the MVD, guided by the developed fitness function.

Main Results:

MHHO demonstrated rapid convergence, typically within ten iterations, significantly outperforming ten recent rival methods.
The approach achieved substantial data reduction, removing 75% of views and decreasing feature size by 66%.
Achieved classification accuracy approaching 100%, with statistical analyses confirming its significant advantage over competitors.

Conclusions:

MHHO is a highly effective and efficient feature selection method for anxiety MVD.
The proposed method significantly enhances classification accuracy and reduces computational load in mental disorder analysis.
Further research can explore the identified anxiety-inducing features and their correlation with comorbid conditions like depression and stress.