Metadata file specification for predictive models
Learn about the available input mapping and outcome categories for your custom artificial intelligence (AI) and machine learning (ML) models. Use these parameters to externally connect and run these models in third-party machine learning services.
JSON file properties for MLaaS models
- objective
- The objective of the model that you want to predict. You can insert any
value, for example,
Churn
. - outcomeType
- The type of outcome that the model predicts. The following values are available:
- BINARY
- Set this value for scoring models that predict two outcome categories.
- CATEGORICAL
- Set this value for categorical models that predict more than two outcome categories.
- CONTINUOUS
- Set this value for continuous models that predict the outcome between a minimum and a maximum value.
- expectedPerformance
- A numeric value that represents the expected predictive performance of
the model. The following values are available:
- 0 to 100
- The range that is available for scoring and categorical models.
- 0 to 1000
- The range that is available for continuous models.
- expectedPerformanceMeasure
- The metric that is used for measuring the expected performance. The
following values are available:
- AUC
- Shows the total predictive performance for scoring models in the Area Under the Curve (AUC) measurement unit. Models with an AUC of 50 provide random outcomes, while models with an AUC of 100 predict the outcome perfectly.
- F-score
- Shows the weighted harmonic mean of precision and recall for categorical models, where precision is the number of correct positive results divided by the number of all positive results returned by the classifier, and recall is the number of correct positive results divided by the number of all relevant samples. The F-score of 1 means perfect precision and recall, while 0 means no precision or recall.
- RMSE
- Shows the root-mean-square error value for continuous models that is calculated as the square root of the average of squared errors. In this measure of predictive power, the difference between the predicted outcomes and the actual outcomes is represented by a number, where 0 means flawless performance.
- framework
- The algorithm that determines the output of the model and converts that
output to a more predictive output format, for example:
- SCIKIT_LEARN
- TENSORFLOW
- XGBOOST
- possibleOutcomes
- The type of outcome that the model predicts. The following values are available:
- Scoring
- The scoring model type declares outcomes as a positive and a negative value, for example, Yes or No.
- Categorical
- In the categorical model type, outcomes can have more than one value which determines the output of the model, for example, categories a, b, c, and d.
- Continuous
- The continuous model type declares the outcome between a minimum and a maximum value, for example, 67.7865 and 88.567.