bf6a52b2-b595-4358-ac4f-24fb41a85c45

R Diabetes Prediction

Overview

PIMA Diabetes demo model using R

Datasets

The dataset required to train or evaluate this model is the PIMA Indians Diabetes dataset available here. This dataset is available in Teradata Vantage and already configured in the demo environment.

Training

{
    "table": "<training dataset>"
}

Evaluation

{
    "table": "<test dataset>"
}

Batch Scoring

 {
     "table": "<score dataset>",
     "predictions": "<ouput predictions dataset>"
 }

Training

The training.R produces the following artifacts

• model.rds (gbm parameters)

Evaluation

Evaluation is also performed in scoring.R by the function evaluate and it returns the following metrics

accuracy: <acc>

Scoring

The scoring.R loads the model and metadata and accepts the dataframe for prediction.

Batch mode

In this example, the values to score are in the table 'PIMA_TEST' at Teradata Vantage. The results are saved in the table 'PIMA_PREDICTIONS'. When batch deploying, this custom values should be specified:

key	value
table	PIMA_TEST
predictions	PIMA_PREDICTIONS

RESTful Sample Request

curl -X POST http://localhost:5000/predict \
        -H "Content-Type: application/json" \
        -d '{
            "data": {
                "ndarray": [[
                        6,
                        148,
                        72,
                        35,
                        0,
                        33.6,
                        0.627,
                        50
                ]],
                "names":[
                    "NumTimesPrg", 
                    "PlGlcConc", 
                    "BloodP", 
                    "SkinThick", 
                    "TwoHourSerIns", 
                    "BMI", 
                    "DiPedFunc", 
                    "Age"
                ]
            }
        }'