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

Algorithm 1

Deep-stacking network flowchart.

	Input: training data .
	Output: ensemble classifier H and risk level P procedure
	Step 1: unstructured data encoding.
	for to m do
	Step 2: standardize data.
	for to m do
	, as equation (1)
	Step 3: multigranularity scanning data based on padding.
	for to m do
	Multigranularity scanning as equation (3)
	Step 4: K-fold cross-validation on the data.
	as equation (4)
	Step 5: learn base-level 1 classifiers.
	for to do
	learn based on
	Step 6: construct new dataset of level 1 predictions and training data.
	for to m do
	where
	Step 7: learn base-level2 classifiers.
	for to do
	learn based on
	Step 8: construct a new dataset of level2 predictions.
	for to m do
	where
	Step 9: learn a metaclassifier.
	learn H based on
	calculate prediction result P as equation (6)
	return H and
	End for and iterative computations.
	End Procedure.