Configure Scorecard Prediction - Platform For AI

Scorecard prediction is a machine learning technique that applies a scorecard model to score new data and predict its future performance or risk. The model is typically generated by a scorecard training component. The scorecard prediction component uses this model to evaluate and score input data, which facilitates decision-making and risk management.

Configure the component

Method 1: Use the UI

In the Designer workflow, add the Scorecard Prediction component and configure its parameters in the pane on the right:

Parameter Type	Parameter	Description
Field settings	Feature columns	Select the original feature columns for prediction. By default, all columns are selected.
	Write to the sink table without transformation	Select columns to append to the prediction result table without any processing, such as ID and target columns.
	Output variable score	Specifies whether to output the score for each feature variable. The final total prediction score is the sum of the intercept's score and all variable scores.
Execution tuning	Number of cores	The number of CPU cores to use. The system allocates this automatically by default.
Execution tuning	Memory per core	The amount of memory for each CPU core. The system allocates this automatically by default.

Method 2: Use a PAI command

You can use a PAI command to configure the parameters for the Scorecard Prediction component. Use the SQL script component to call the PAI command. For more information, see SQL Script.

pai -name=lm_predict
    -project=algo_public
    -DinputFeatureTableName=input_data_table
    -DinputModelTableName=input_model_table
    -DmetaColNames=sample_key,label
    -DfeatureColNames=fea1,fea2
    -DoutputTableName=output_score_table

Parameter	Required	Default value	Description
inputFeatureTableName	Yes	None	The input feature data table.
inputFeatureTablePartitions	No	The entire table	The partitions to select from the input feature table.
inputModelTableName	Yes	None	The input model table.
featureColNames	No	All columns	The feature columns to select from the input table.
metaColNames	No	None	Data columns that are not transformed. The selected columns are output as is. Specify columns such as the label and sample_id here.
outputFeatureScore	No	false	Specifies whether to output variable scores in the prediction result. Valid values: true: Outputs variable scores. false: Does not output variable scores.
outputTableName	Yes	None	The output prediction result table.
lifecycle	No	None	The lifecycle of the output table.
coreNum	No	Calculated automatically	The number of cores.
memSizePerCore	No	Calculated automatically	The memory size in MB.

Component output

The Scorecard Prediction component outputs a scoring table, as shown in the following example: Scoring table example The churn column is added to the result table as is and is not related to the prediction result. The other three columns are prediction result columns, which are described in the following table.

Column name	Column type	Column description
prediction_score	DOUBLE	The prediction score column. In a linear model, this is the result of multiplying the feature values by the model weight values and then summing them. In a scorecard model, if the model performs score transformation, this column outputs the transformed score.
prediction_prob	DOUBLE	In a binary classification scenario, this is the predicted probability value of the positive sample. This value is obtained by applying a Sigmoid transformation to the original score (before score transformation).
prediction_detail	STRING	The probability values for each class, described in JSON format. 0 represents the negative class, and 1 represents the positive class. For example: {“0”:0.1813110520,”1”:0.8186889480}.