Mixed Script Identification Using Automated DNN Hyperparameter Optimization

Table 8

Test results for Bi-GRU model.

Dataset	RMSE	nRMSE (%)	MBE	MAE	R^2
Eng-Roman Urdu	355.9	7.70	12.5	158	0.89
Saraiki-Hindi	380.7	9.10	3.5	174	0.91
Bengali-Hindi	410.6	6.73	6.4	191	0.87
Eng-Bengali	392.4	4.35	9.76	127	0.87
Saraiki-English	282.8	3.90	5.76	141	0.81
Saraiki-Roman Urdu	365.9	6.34	6.56	136	0.8
Saraiki-Bengali	401.2	6.90	8.54	189	0.86
Eng-Bengali-Saraiki-Hindi-Roman Urdu	673.6	11.50	9.32	174	0.75