Comparisons of Prediction Models of Myofascial Pain Control after Dry Needling: A Prospective Study

Table 4

Comparison of multiple linear regression (MLR), support vector machine (SVM), and artificial neural network (ANN) models in predicting Brief Pain Inventory (BPI) scores.

Indices	Models	Training set (𝐴)	Testing set (𝐵)	Change rate^#
Worst pain
MSE	MLR	22.41	24.37	8.7%
	SVM	16.05	14.52	10.5%
	ANN	15.02	12.63	20.3%
MAPE	MLR	8.5%	8.1%	—
	SVM	5.9%	5.1%	—
	ANN	4.4%	4.5%	—
Average pain
MSE	MLR	19.19	17.84	7.6%
	SVM	13.93	12.86	8.3%
	ANN	13.26	11.56	14.7%
MAPE	MLR	6.4%	6.2%	—
	SVM	5.5%	5.9%	—
	ANN	4.0%	4.1%	—
Present pain
MSE	MLR	17.68	18.82	6.1%
	SVM	12.06	13.01	7.3%
	ANN	10.31	11.16	7.6%
MAPE	MLR	6.9%	6.9%	—
	SVM	5.7%	5.0%	—
	ANN	4.6%	4.4%	—
Aggregated pain interference
MSE	MLR	14.83	14.28	3.9%
	SVM	11.06	10.18	8.6%
	ANN	8.13	8.91	8.8%
MAPE	MLR	5.6%	5.4%	—
	SVM	4.5%	4.7%	—
	ANN	3.4%	3.4%	—

MSE: mean square error, MAPE: mean absolute percentage error.
#Change rate = |(𝐵−𝐴)/(𝐴)|×100%.