A comparison of regression analysis for predicting the daily number of anxiety-related outpatient visits with different time series data mining

Abstract

This study aimed to develop and evaluate different models to forecast the daily number of anxiety-relatedpatients seeking to visit the outpatient department in Prasrimahabhodi Psychiatric Hospital. The authors developed and tested four different models of outpatient visits using total daily counts of anxiety-related patient visits to outpatient at Prasrimahabhodi Psychiatric Hospital, Thailand from January 2011 to December 2013.Multi-Layer Perceptron Regression (MLPR), Radial basis function Regression (RBFR), and Support Vector Regression (SVR) were compared with the traditional statistical tool of Linear Regression (LR). The sliding window method was used to prepare the dataset for the number of anxiety-related outpatient visits forecasting process. The performances of the models were compared in terms of the mean absolute error (MAE) and root mean square error (RMSE). The performance comparison showed that the SVR exhibited a slightly better performance. The SVR also showed highly stable. The outcome of the study can be of use for planning staff arrangement and material resources distribution.