Example 36.1. Classification building example

The script below shows how to build a model using the LVQ algorithm.

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name)
save('physical_data', pd) 
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- LEARNING - 1st stage --- ---
map_height = 12 
map_width = 12

# --- algorithm settings for SOM --- 
som_cfs = ClusteringFunctionSettings() 
som_cfs.logicalData = ld 
som_set = KohonenClusteringSettings()
som_set.maxNumberOfIterations = 30 
som_set.height = map_height 
som_set.width = map_width
som_cfs.algorithmSettings = som_set
save('SOM_cf_set', som_cfs)

# --- SOM model building --- 
SOM_model_name = 'SOM_model_' + input_data_name
som_mbt = MiningBuildTask('physical_data', 'SOM_cf_set', SOM_model_name)
save('SOM_mbt', som_mbt)
execute('SOM_mbt')

# --- --- LEARNING - 2nd stage --- ---

# --- algorithm settings for LVQ --- 
lvq_cfs = ClassificationFunctionSettings()
lvq_cfs.logicalData = ld
lvq_cfs.getAttributeUsageSet().getAttribute('Class').setUsage(UsageOption.target)
lvq_set = KohonenClassificationSettings(ClassificationLearningAlgorithmType.lvq21)
lvq_set.height = map_height 
lvq_set.width = map_width

lvq_cfs.algorithmSettings = lvq_set
save('LVQ_cf_set', lvq_cfs)

# --- LVQ model building --- 
LVQ_model_name = 'LVQ_model_' + input_data_name
lvq_mbt = MiningBuildTask('physical_data', 'LVQ_cf_set', LVQ_model_name )
lvq_mbt.inputModelName = SOM_model_name 
save('LVQ_mbt', lvq_mbt)
execute('LVQ_mbt')
    

Example 36.2. Clustering building example

This example shows how to build a clustering model using the Kohonen SOM algorithm.

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name)
save('physical_data', pd) 
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- LEARNING --- ---

# --- algorithm settings for SOM --- 
som_cfs =ClusteringFunctionSettings() 
som_cfs.logicalData =ld
som_cfs.getAttributeUsageSet().getAttribute('Class').setUsage(UsageOption.inactive)

som_set = KohonenClusteringSettings()
som_set.batchLearning = FALSE
som_set.beginLearningRate = 0.3
som_set.beginNeighbourhoodRate = 2
som_set.dataRandomization = TRUE
som_set.setDistanceFunction(ClusteringDistanceFunction.euclidean)
som_set.endLearningRate = 0.1
som_set.endNeighbourhoodRate = 0 
som_set.height = 15
som_set.maxNumberOfIterations = 150
som_set.minErrorTolerance = 0.01
som_set.setModificationFunction(ModificationFunction.standard)
som_set.noiseVariance = -1.0
som_set.setSearchType(SearchType.global)
som_set.torus = FALSE 
som_set.validationWindow = 10
som_set.width = 15 
som_cfs.algorithmSettings =som_set
save('SOM_cf_set', som_cfs)

# --- SOM model building --- 
SOM_model_name = 'SOM_model_' + input_data_name 
som_mbt = MiningBuildTask('physical_data','SOM_cf_set',SOM_model_name)
save('SOM_mbt', som_mbt)
execute('SOM_mbt')
    

Example 36.3. Calculate statistics example

This example shows how to calculate the statistics for SOM and LVQ models.

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name)
save('physical_data', pd) 
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- STATISTICS --- ---
LVQ_model_name = 'LVQ_model_' + input_data_name

# --- statistics for LVQ model --- 
cmst = ComputeModelStatisticsTask()
cmst.modelName = LVQ_model_name
cmst.modelStatResultName = LVQ_model_name + '_statistics'
cmst.testDataName = 'physical_data'
save('MAP_cmst',cmst) 
execute('MAP_cmst')
    

Example 36.4. Classification applying example

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name)
save('physical_data', pd)
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- model application --- ---
LVQ_model_name = 'LVQ_model_' + input_data_name

# --- using the model (LVQ) --- 
lvq_ao_pd = PhysicalData('LVQ_apply_output_for_' + input_data_name)
save('LVQ_apply_output',lvq_ao_pd)

lvq_mat = MiningApplyTask()
lvq_mat.modelName = LVQ_model_name
lvq_mat.sourceDataName = 'physical_data'
lvq_mat.targetDataName = 'LVQ_apply_output'
lvq_mat.replaceExistingData = TRUE
lvq_mat.directMapping.add(ApplySourceItem('Class','real_target_value'))

lvq_cao = ClassificationApplyOutput()
lvq_cao.item.add(ClassificationRankItem('predicted_target',ClassificationOutputType.predictedCategory,0))
lvq_cao.item.add(ClassificationRankItem('prediction_probability',ClassificationOutputType.probability,0))
lvq_mat.applyOutput = lvq_cao

save('LVQ_apply_task', lvq_mat)
execute('LVQ_apply_task')
    

Example 36.5. Clustering applying example

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name)
save('physical_data', pd)
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- model application --- ---
SOM_model_name = 'SOM_model_' + input_data_name

# --- using the model (SOM) --- 
som_ao_pd = PhysicalData('SOM_apply_output_for_' + input_data_name )
save('SOM_apply_output', som_ao_pd)

som_mat = MiningApplyTask() 
som_mat.modelName = SOM_model_name 
som_mat.sourceDataName = 'physical_data' 
som_mat.targetDataName = 'SOM_apply_output' 
som_mat.replaceExistingData = TRUE

som_cao = ClusteringApplyOutput()
som_cao.item.add(ClusteringRankItem('cluster_id',ClusteringOutputType.clusterId ,0))
som_cao.item.add(ClusteringRankItem('cluster_0_distance',ClusteringOutputType.distance,0))
som_cao.item.add(ClusteringRankItem('cluster_1_distance',ClusteringOutputType.distance,1))
som_cao.item.add(ClusteringRankItem('cluster_2_distance',ClusteringOutputType.distance,2))
som_cao.item.add(ClusteringRankItem('cluster_3_distance',ClusteringOutputType.distance,3))
som_mat.applyOutput = som_cao

save('SOM_apply_task', som_mat)
execute('SOM_apply_task')
    

Example 36.6. Testing example

# --- --- data preparation --- ---
input_data_name = 'iris'

pd = PhysicalData(input_data_name) 
save('physical_data',pd)
ld = LogicalData(pd)
save('logical_data', ld)

# --- --- model testing --- ---
LVQ_model_name = 'LVQ_model_' + input_data_name

lvq_ctt = ClassificationTestTask()
lvq_ctt.positiveTargetValue = "Iris-versicolor"
lvq_ctt.testDataTargetAttributeName = "Class"
lvq_ctt.modelName = LVQ_model_name
lvq_ctt.testDataName = 'physical_data'
lvq_ctt.testResultName = 'results_' + LVQ_model_name
save('LVQ_ctt',lvq_ctt) 
execute('LVQ_ctt')