Deep-Stacking Network Approach by Multisource Data Mining for Hazardous Risk Identification in IoT-Based Intelligent Food Management Systems

Table 3

Identification result of the risk level in comparative methods.

Models	Accuracy (%)		Model parameters (megabyte)
	Architecture-1	Architecture-2	Architecture-1	Architecture-2
Logistic regression [19]	59.21	69.35	41.31	43.25
K-nearest neighbour [35]	66.32	69.01	50.65	61.36
Support vector machine [20]	66.70	72.60	47.33	58.97
Extra trees [18]	71.88	72.83	63.57	70.65
Gradient boosting [22]	78.94	81.29	61.37	66.91
Random forest [26]	80.51	83.93	70.18	81.32
Decision tree [17]	83.97	85.45	72.34	79.25
AlexNet [26]	70.62	71.32	218.96	223.54
VGG [28]	72.88	73.94	320.78	365.12
GoogLeNet [27]	85.18	87.98	425.89	478.23
ResNet [29]	87.26	89.68	507.34	528.67
Deep forests [32]	90.73	92.47	385.32	419.11
Proposed DSN	94.88	97.62	202.53	211.26