This tutorial show how to apply an example classification model in AdvancedMiner. All the quoted names are example names of created objects.
To apply a model, it is required to:
-
have a model object in the repository
- the process of creating a model is described in the Model Building Guide tutorial.
- have a table in the database with data for scoring.
Assumption: in the following example, it is assumed that the user has only a single model object in the MR repositiory, built during a certain model building process.
During the model applying process, the following MR objects are used:
- PhysicalData - two object of this type: the first one describes data for scoring, the second one is connected with the output table,
- MiningApplyTask - an object controlling the whole model applying process
- model - the result of the model building process
Model applying task consists of the following steps:
-
step 1:
- create PhysicalData ('german_credit_pd')
- we select the data for scoring: 'german_credit'
-
step 2:
- create PhysicalData ('apply_results_pd')
- we set a new name for the output table: 'apply_results'
- step 3:
- create MiningApplyTask ('MiningApplyTask')
-
we define that we are going to apply a model
-
step 3a:
- add ClassificationApplyOutput
-
we choose the type of output
information: in this case for
a classification model
-
step 3a-i:
- item: add element
- we choose ClassificationCategoryItem and we define the output variable named 'score' as the probability for the category 'bad'.
-
step 3a-i:
-
step 3b:
- add the model
- we choose the model to apply: 'MODEL'
-
step 3c:
- add sourceData
- we choose 'german_credit_pd' for PhysicalData of the data for scoring
-
step 3d:
- add targetData
- we choose 'apply_results_pd' for PhysicalData of the output data
-
step 3e:
- directMapping: add element
- we define the output variable named 'real_target_value' as a copy of the 'Class' variable from input data
-
step 3a:
- step 4:
- execute task 'MiningApplyTask'
Note
- Regardless of the model type, the scheme described above remains the same and constists of the same steps. The differences may lie in how the details of the output columns are defined, as the particular set of available items depend on the model type.
If all the required objects were set up correctly, then after executing MiningApplyTask an output table with two columns should appear in the database ( see the movie ).
The steps described above can be also executed in a script.
Example 7.1. ModelApplying
# MODEL APPLYING PROCESS
# -- we assume that a model named 'MODEL' has already been build and it exists in the repository
# step 1 -- create and save PhysicalData in MR
# -- we choose the data for scoring: 'german_credit'
PD = PhysicalData('german_credit')
save('german_credit_pd',PD)
# step 2 -- create and save PhysicalData in MR
# -- we set a new name for the apply output table
PD = PhysicalData('apply_results')
save('apply_results_pd',PD)
# step 3 -- create MiningApplyTask
# -- we define that we are going to apply a model
MAT = MiningApplyTask()
# step 3a -- MiningApplyTask: add applyOutput -->> ClassificationApplyOutput
# -- we choose the type of output information: in this case for a classification model
CAO = ClassificationApplyOutput()
# step 3a-i-- MiningApplyTask -> ClassificationApplyOutput -> item: add element -->> ClassificationCategoryItem
# -- we define the output variable 'score' as the probability for the category 'bad'
CCI_score = ClassificationCategoryItem('score' , ClassificationOutputType.probability , 'bad')
CAO.item.add ( CCI_score )
# conect ClassificationApplyOutput to MiningApplyTask
MAT.setApplyOutput(CAO)
# step 3b -- MiningApplyTask: add model
# -- we choose the model to apply
MAT.setModelName('MODEL')
# step 3c -- MiningApplyTask: add sourceData
# -- we choose the data for scoring
MAT.setSourceDataName('german_credit_pd')
# step 3d -- MiningApplyTask: add targetData
# -- we choose the output data
MAT.setTargetDataName('apply_results_pd')
# step 3e -- MiningApplyTask -> directMapping: add element
# -- we define an output variable named 'real_target_value' as a copy of the 'Class' variable from input data
ASI_real_target_value = ApplySourceItem('Class','real_target_value')
MAT.directMapping.add( ASI_real_target_value )
# steps 3,3a,3a-1,3b,3c,3d,3e -- save MiningApplyTask in MR
save('MiningApplyTask',MAT)
# step 4 -- execute task 'MiningApplyTask'
execute('MiningApplyTask')
print "Done..."
Output:
Done...