Supervised learning creates a machine learning model by analyzing a labeled dataset to learn patterns. This means that the dataset has values associated with the column (the label) that the machine learning model eventually generates predictions for. The model is able to to predict labels based on the features of the dataset. For example, a census and income dataset may have features such as age, education, occupation, and country that you can use to predict the income of an individual (the label). The income label in this dataset already has values that the machine learning maodel uses for training.

Once a machine learning model is trained, it can be used on unseen data, where the label is unknown, to make predictions. In a business setting, predictive models have a variety of possible applications such as predicting customer churn, approving or rejecting credit applications, predicting customer wait times, and so on.

See Labeled Data and Unlabeled Data to learn more.

MySQL HeatWave AutoML supports unsupervised learning for forecasting, anomaly detection and topic modeling models. This type of learning requires no labeled data. This means that the column (the label) the machine learning model eventually generates predictions for has no values in the dataset for training. For example, a dataset of credit card transactions that you use for anomaly detection has a column indicating if the transaction is anomalous or normal, but the column has no data (unlabeled). See Generate Forecasts, Detect Anomalies, and Generate Topic Modeling to learn more.

MySQL 9.0.1-u1 introduces support for semi-supervised learning when running anomaly detection. This type of machine learning algorithm uses a specific set of labeled data along with unlabeled data to detect anomalies. The dataset for this type of model must have a column whose only allowed values are 0 (normal), 1, (anomalous), and NULL (unlabeled). All rows in the dataset are used to train the unsupervised component, while the rows with a value different than NULL are used to train the supervised component. See Detect Anomalies and Anomaly Detection Model Types to learn more.