Predicting Parkinson’s Disease Progression: Evaluation of Ensemble Methods in Machine Learning

Table 3

SOM and EM ensembles by CSPA, HGPA, and majority voting for UPDRS prediction.

Ensemble size (SOM)	Ensemble technique	RMSE	MAE	IA	PA
Motor-UPDRS
2(SOM2 × 3 + SOM2 × 4)	CSPA	0.5980	0.4437	0.9208	0.9155	0.8999
	HGPA	0.5960	0.4416	0.9238	0.9200	0.9019
3(SOM2 × 4 + SOM3 × 3 + SOM3 × 4)	Majority voting	0.5720	0.4152	0.9306	0.9265	0.9078
	CSPA	0.5620	0.4150	0.9309	0.9266	0.9085
	HGPA	0.5540	0.4116	0.9335	0.9277	0.9139
4(SOM2 × 3 + SOM2 × 4 + SOM3 × 3 + SOM3 × 4)	CSPA	0.5772	0.4287	0.9283	0.9261	0.9067
	HGPA	0.5756	0.4286	0.9287	0.9263	0.9070
Total-UPDRS
2(SOM2 × 3 + SOM2 × 4)	CSPA	0.6053	0.4463	0.9141	0.9101	0.8872
	HGPA	0.6016	0.4432	0.9166	0.9104	0.8928
3(SOM2 × 4 + SOM3 × 3 + SOM3 × 4)	Majority voting	0.5756	0.4269	0.9207	0.9173	0.9043
	CSPA	0.5700	0.4250	0.9230	0.9203	0.9043
	HGPA	0.5565	0.4179	0.9289	0.9240	0.9058
4(SOM2 × 3 + SOM2 × 4 + SOM3 × 3 + SOM3 × 4)	CSPA	0.5853	0.4395	0.9182	0.9138	0.8983
	HGPA	0.5799	0.4376	0.9193	0.9140	0.9026
Ensemble size (EM)	Ensemble technique	RMSE	MAE	IA	PA
Motor-UPDRS
2(k = 8,10)	CSPA	0.6122	0.4557	0.9103	0.9080	0.8904
	HGPA	0.5974	0.4521	0.9118	0.9092	0.8963
3(k = 8,10,12)	Majority voting	0.5897	0.4414	0.9149	0.9112	0.8998
	CSPA	0.5795	0.4397	0.9156	0.9117	0.9010
	HGPA	0.5789	0.4287	0.9173	0.9140	0.9028
4(k = 8,10,12,13)	CSPA	0.5623	0.4171	0.9257	0.9198	0.9049
	HGPA	0.5594	0.4165	0.9319	0.9254	0.9130
Total-UPDRS
2(k = 8,10)	CSPA	0.6137	0.4574	0.9085	0.9049	0.8837
	HGPA	0.6086	0.4531	0.9092	0.9071	0.8848
3(k = 8,10,12)	Majority voting	0.5921	0.4480	0.9120	0.9081	0.8872
	CSPA	0.5912	0.4466	0.9135	0.9100	0.8904
	HGPA	0.5846	0.4398	0.9173	0.9101	0.8933
4(k = 8,10,12,13)	CSPA	0.5788	0.4370	0.9216	0.9180	0.8957
	HGPA	0.5665	0.4186	0.9281	0.9209	0.9018