Prediction details are XML strings that provide information about the score. Details are available for all types of scoring: clustering, feature extraction, classification, regression, and anomaly detection. Details are available whether scoring is dynamic or the result of model apply.
The details functions, CLUSTER_DETAILS, FEATURE_DETAILS, and PREDICTION_DETAILS return the actual value of attributes used for scoring and the relative importance of the attributes in determining the score. By default, the functions return the five most important attributes in descending order of importance.