Table 1 Performance comparison of different lifelong local predictors for predicting total virus and specific viral genera across the aerobic treatment process

Model	Set	Total Virus				AdV				PMMoV
Model	Set	R²	RMSE	MAE	SMAPE	R²	RMSE	MAE	SMAPE	R²	RMSE	MAE	SMAPE
PLS	Train	0.514	0.263	0.198	2.139	0.814	0.304	0.226	2.092	0.754	0.426	0.322	2.282
	Test	0.602	0.231	0.171	1.854	0.841	0.276	0.202	1.876	0.770	0.436	0.336	1.790
	Adaptation (aerobic)	0.970	0.017	0.012	0.203	0.910	0.235	0.181	0.259	0.938	0.201	0.149	0.255
XGB	Train	0.971	0.064	0.044	0.479	0.989	0.073	0.052	0.805	0.988	0.093	0.066	0.509
	Test	0.726	0.191	0.132	1.422	0.864	0.255	0.187	1.765	0.851	0.351	0.252	1.123
	Adaptation (aerobic)	0.629	0.065	0.050	0.613	0.659	0.470	0.358	0.624	0.673	0.551	0.408	0.613
CatBoost	Train	0.820	0.160	0.115	1.237	0.920	0.199	0.147	1.140	0.918	0.246	0.186	1.265
	Test	0.723	0.192	0.136	1.464	0.855	0.347	0.261	1.368	0.853	0.347	0.261	1.125
	Adaptation (aerobic)	0.719	0.057	0.042	0.523	0.741	0.412	0.313	0.572	0.732	0.505	0.362	0.563
GRU	Train	0.863	0.140	0.093	0.973	0.925	0.193	0.129	0.938	0.944	0.203	0.138	0.941
	Test	0.832	0.150	0.096	1.002	0.929	0.185	0.127	0.934	0.943	0.217	0.151	1.047
	Adaptation (Aerobic)	0.974	0.016	0.012	0.214	0.960	0.148	0.108	0.207	0.948	0.184	0.132	0.307
LSTM	Train	0.850	0.146	0.094	0.991	0.925	0.193	0.129	0.958	0.944	0.203	0.138	1.001
	Test	0.831	0.150	0.095	1.028	0.929	0.185	0.127	0.896	0.943	0.217	0.151	0.947
	Adaptation (aerobic)	0.973	0.015	0.011	0.208	0.960	0.148	0.108	0.207	0.947	0.184	0.131	0.285

The best model adaptation performance results for predicting total virus and specific viral genera (i.e., AdV and PMMoV) across the aerobic treatment process, with a higher coefficient of determination R² value and a lower (RMSE, MAE, and SMAPE) value between the local predictors (PLS, XGB, CatBoost, GRU, and LSTM), are highlighted in bold.

Quick links

Search