Random forests based monitoring of human larynx using questionnaire data

• Published in: Expert systems with applications 2012-04, Vol.39 (5), p.5506-5512
• Main Authors: Bacauskiene, M., Verikas, A., Gelzinis, A., Vegiene, A.
• Format: Article
• Language: English
• Subjects: Accuracy; Classification; Classifier; Data proximity; Forests; Human larynx; Larynx; Mathematical analysis; Mathematical models; Monitoring; Radio frequencies; Random forests; Variable importance; Variable selection
• ISSN: 0957-4174; 1873-6793; 1873-6793
• DOI: 10.1016/j.eswa.2011.11.070

This paper is concerned with soft computing techniques-based noninvasive monitoring of human larynx using subject’s questionnaire data. By applying random forests (RF), questionnaire data are categorized into a healthy class and several classes of disorders including: cancerous, noncancerous, diffuse, nodular, paralysis, and an overall pathological class. The most important questionnaire statements are determined using RF variable importance evaluations. To explore data represented by variables used by RF, the t-distributed stochastic neighbor embedding ( t-SNE) and the multidimensional scaling (MDS) are applied to the RF data proximity matrix. When testing the developed tools on a set of data collected from 109 subjects, the 100% classification accuracy was obtained on unseen data in binary classification into the healthy and pathological classes. The accuracy of 80.7% was achieved when classifying the data into the healthy, cancerous, noncancerous classes. The t-SNE and MDS mapping techniques applied allow obtaining two-dimensional maps of data and facilitate data exploration aimed at identifying subjects belonging to a “risk group”. It is expected that the developed tools will be of great help in preventive health care in laryngology.