Forecasting the number of deaths from cerebrovascular diseases in Thailand using grey systems theory

Abstract

Due to uncertainties in life, a forecasting system that can accept variables and generate an acceptable predictive function needs to be developed.  The grey system theory works with poor, incomplete or uncertain past data in time series forecasting.  According the Public Health Statistics, the number of deaths from cerebrovascular diseases in Thailand were S-shaped with an overall upward trend from 1996 to 2015; with deaths rising from 6,300 to 8,200 during 1996-2000 (V-shape), 11,309 up to 19,265 and down to 15,719 during 2001 to 2005(/\-shape ), rose from 12,921 to 17,540 during 2006 to 2010, and rose again from 19,283 to 28,146 in 2011 to 2015.  The data are arranged in 4 sets from the past to the most current year 2015: there are 5, 10, 15, and 20 years as the input to 5 grey models: Grey Model First Order One Variable (GM(1,1)), expanded forms of GM(1,1) (GM(1,1)E), and expanded forms of GM(1,1) with residuals correction (GM(1,1)E&C), the grey Verhulst model (VGM), and the grey Verhulst model with improvement (VGMI).  From the 20 solutions from 4 sets of input and 5 models, the forecasts from GM(1,1)E&C show minimum errors with high correlation efficiencies.  The different shape and number of past data also affect the forecast values.  The forecast number of deaths from cerebrovascular in 2016 will be 25,991 persons, if the S-shape in the past will be repeated.