Model builder results

When you build and deploy a model with IBM Watson Studio you can choose for Model Builder to automatically match your data with a classifier and algorithm or choose them yourself. After you build and deploy a model, you can test the model and view the results.

Sample results

This example shows a prediction for whether a customer of a sporting goods store is likely to purchase a tent, based on given data such as age, gender and spending history. You can see that for the test data supplied, the prediction is False; the customer is not likely to purchase a tent.

Testing a model

Reviewing the results

To examine the model results, click the icon to View raw output to view the model results as JSON code. You will see output similar to this:

  "fields": [


Understanding the results

In the resulting code you can see the columns you defined, such as GENDER, AGE, and PURCHASE_AMOUNT, plus some output columns automatically generated by the Model Builder. The generated columns differ depending on whether you are building a classifier type of model or a regression model.

This table lists results automatically generated by Model Builder according to model type.

Table 1. Model results generated by model type
Model type Automatically generated results
  • features
  • rawPrediction
  • Probability
  • Prediction
  • nodeADP_class
  • nodeADP_classes
Multiclass classifier
  • features
  • rawPrediction
  • Probability
  • Prediction
  • nodeADP_class
  • nodeADP_classes
  • features
  • Prediction

Refer to this table to understand what the generated output represents for your model.

Table 2. Description of results generated by Model Builder
Column Description
features A property, such as a column name, of your training data.
rawPrediction Raw prediction for each possible label. The meaning of a "raw" prediction may vary between algorithms, but it intuitively gives a measure of confidence in each possible label, where a larger number represents a more confident score.
Probability The Probability is the conditional probability for each class, given the raw prediction. The actual calculation depends on which Classifier you are using.
Prediction Prediction is the argument where the array probability takes its maximum value, and it gives the most probable label. In particular, for classification models, you get the predicted probability of each class (that is, class conditional probabilities); for regression, you get the biased sample variance of prediction.
nodeADP_class The nodeADP_class column indicates the prediction of the row based on your model.