Estimation the Health Literacy Score of Syrian Immigrants Using Support Vector Regression, Decision Tree, Extreme Learning Machine, and Deep Autoencoder

Abstract

Health literacy is defined as social policy. This study aims to determine the health literacy level of Syrian immigrants living in Elazig, Turkey and its relevant factors and to predict it with artificial intelligence methods. The minimum sample size was calculated as 373 by using the Epi Info program. The questionnaire included socio-demographic information and the health literacy scale (HLS) EU-16. Means were presented with standard deviation (mean±SD), and p<0.05. was considered statistically significant. Besides performance evaluation, Support Vector Regression (SVR), Decision Tree (DT), Extreme Learning Machines (ELM), and Autoencoder Regression (AR) were used to estimate the HLS score. The mean age of 404 Syrian immigrants who participated in the study was 32.64±10.97 years. They had been in Turkey for an average of 5.00±2.267 years. The mean total HLS score was 51.55±12.915. The HLS value of the Syrian immigrants was calculated using four different regression methods based on machine learning according to three basic criteria (Age, Duration of Stay in Turkey, and Monthly Income). Autoencoder, ELM, and DT regression methods indicated that the average estimation success was above 91% according to the R2 performance criterion, and the highest success rate of 98% was achieved with the autoencoder method.