For model testing based on scoring result

Read(112) Label: scoring, modelperformance test, json,

If the data to be scored in batch contains the target variable, we can get model performance information according to the scoring result. Same as explained in the previous section, we can get the information in JSON format through the YModel’s special interface when the YModel is integrated into a third-party program or when we use Yimming as an external library.

Parameters

 

Parameters and their descriptions are the same as those for model performance.

Example

{

"Model-Performance":

{

"KS":0.37160852713178294,

"ROC-Data":["{

"1-specificity":"0.0",

"sensitivity":"0.020833333333333332"

},

// Next set of data

]

"Lift-Recall-Data":[

{

"CumulativeRecall":"0.10416666666666667",

"Decile":"1",

"CumulativeLift":"2.147435897435898",

"Recall":"0.10416666666666667",

"Lift":"2.147435897435898"

},

// Next set of data

]

"GINI":0.4704457364341086,

"Accu-Data:[

{

"precision":"0.3735408560311284",

"recall":"1.0","accuracy":"0.39925373134328357",

"threshold":"0.05000000074505806"

},

// Next set of data

]

"AUC":0.7352228682170543

}

}