Short-Term Prediction of Traffic State for a Rural Road Applying Ensemble Learning Process

Table 7

Confusion matrices of predictions for the test dataset.

True prediction	A	B	C
Model	LSTM-all
A	151	151	3
B	19	542	140
C	0	0	1
Model	SVM-all
A	136	81	0
B	34	610	131
C	0	2	13
Model	KNN-all
A	144	97	1
B	26	585	125
C	0	11	18
Model	RF-all
A	150	78	0
B	20	614	133
C	0	1	11
Model	Voting to a better state
A	149	86	1
B	21	607	142
C	0	0	1
Model	Best state
A	162	175	3
B	8	518	140
C	0	0	1
Model	OL
A	141	36	0
B	29	649	118
C	0	8	26
Model	LSTM-GA
A	137	119	1
B	33	574	142
C	0	0	1
Model	SVM-GA
A	143	88	1
B	27	605	142
C	0	0	1
Model	KNN-GA
A	144	90	1
B	26	594	140
C	0	9	3
Model	RF-GA
A	151	76	0
B	19	613	134
C	0	4	10
Model	Voting to a worse state
A	142	76	1
B	28	617	140
C	0	0	3
Model	Worst state
A	116	26	0
B	54	645	116
C	0	22	28