Forecasting Model on Mental Illness Patients with High Risk for Violence (SMI-V) Not Causing Repeated Violence, Buriram Province

Abstract

Introduction: Forecasting the number of psychiatric patients provides useful information for decision-making in preparing to develop service approaches and allocate appropriate budgets to healthcare facilities.

Methods:  The data were drawn from the HDC reporting system of Buriram Provincial Public Health Office using number of psychiatric patients depend on the time series from budget years 2013 to 2022, totaling 120 values. The data were divided into 2 sets with the first set from October 2012 to September 2021, consisting of 108 values. The forecasting study methods includes 6 ways using Excel program. The second set from October 2021 to September 2022, consisting of 12 values, using  the  accuracy of forecast values to compare, by the mean absolute deviation (MAD) and the lowest level of mean absolute percentage error (MAPE).

Results: It was founded that the most suitable model was the double exponential soothing: Holt’s two-parameter (MAPE=11.07), followed by the straight-line moving average method (MAPE=16.69) and Winter’s three-parameter trend & seasonality method (Winters) (MAPE=19.53).

Discussion: among all 6 ways of forecasting models, it was founded that the double exponential soothing: Holt’s two-parameter was the most suitable model for  this study with the lowest mean absolute percentage error of 11.07.

Conclusion: the double exponential soothing: Holt’s two-parameter showed the decrease in mental illness patients with the increasing trend of patients from June to September.