Misfire Detection in Spark Ignition Engine Using Transfer Learning

Table 4

Performance of pretrained models for various train-test split ratios

Pretrained model	Classification accuracy for train-test split ratio (%)					Overall accuracy (%)
	0.60	0.70	0.75	0.80	0.85
AlexNet	89.00	90.70	88.80	90.00	90.30	89.76
VGG-16	95.00	92.70	96.80	97.00	97.30	95.76
GoogLeNet	92.00	88.70	85.60	96.00	92.00	90.86
ResNet-50	84.00	86.70	93.60	89.00	84.00	87.46

The values highlighted in bold signify the best classification accuracy delivered by the pretrained model for the experimented train-test split ratio. While considering the last column, the best pretrained network with highest classification accuracy value is highlighted in bold.