Research Article
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. |
|