Automatic Verbal Autopsy Classification Using Multi-Class Linear Discriminant Analysis and Recursive Feature Elimination

Abstract

Verbal autopsy, also referred to as postmortem is a scientific technique via which persons precise purpose of demise may be recognized. This procedure is carried out beneath supervision of medical experts who generally follows the guidelines supplied by using World Health Organization (WHO). Poor countries, wherein the clinical centers are very uncommon loss of life of human beings can manifest normally outside of the hospital. Understanding the precise cause of demise of humans is complicated and fuzzy. Computerized Verbal autopsy is automatic system via which we can become aware of the motive of loss of life of any man or woman automatically without repeating the clinical technique. In this paper we used exclusive popular datasets which might be regular and published by means of world Health organization. These datasets normally which includes binary values so we used the Linear discriminant analysis to classifying the exact cause of death. To get right of entry to the overall performance of our proposed model and previous models like InterVA, Tariff-4 and InsilicoVA we used metrics like Sensitivity, Chance Corrected Concordance (CCC) and cause-specific mortality fraction (CSMF).  Experimental effects indicates that proposed model is better than the previous models.