Abstract: Various embodiments of the present disclosure describe data evaluation techniques that leverage a graph-based machine learning model to evaluate a knowledge graph. The techniques include using a target graph model to generate a predictive representation for a graph node of a graph training dataset. The techniques include using a feature prediction model to generate predicted feature values for the graph node based on the predictive representation. The techniques include generating a data evaluation score for the graph training dataset based on the predicted feature values. The techniques include using the target graph model to generate a predictive output for the graph node based on the predictive representation and then generating an evaluation output for the target graph model based on the evaluation score and the predictive output.

Abstract: Various embodiments of the present disclosure describe holistic machine learning model evaluation techniques. The techniques include determining a holistic evaluation vector for a target machine learning model based on a plurality of evaluation scores for the target machine learning model. The plurality of evaluation scores may include a data evaluation score corresponding to a training dataset for the target machine learning model, a model evaluation score corresponding to one or more performance metrics for the target machine learning model, and a decision evaluation score corresponding to an output class of the target machine learning model. A holistic evaluation score for the target machine learning model may be determined from the holistic evaluation vector or a plurality of evaluation scores. An informed evaluation output is provided based on the holistic vector or score.

Abstract: Various embodiments of the present disclosure describe machine learning monitoring and retraining techniques for automatically triggering model retraining based on evaluation scores. The techniques include receiving a request to process an input data object with a target machine learning model that is previously trained using an at least partially synthetic training dataset. The techniques include identifying a synthetic data object from the training dataset that corresponds to the input data object and, in response, modifying a holistic evaluation score for the model, initiating the performance of a labeling process for assigning a ground truth label to the input data object, and augmenting a supplemental training dataset with the input data object and the ground truth label. In the event that the holistic evaluation score decreased beyond a threshold, the model may be retrained with the supplemental training dataset.