Choose this use-case if you wish to develop a model that makes predictions using new data as it comes in. Processing in real-time is especially useful in fast-paced environments, where quick decision-making, proactive response, and making predictions on fresh data are most relevant. Given a dataset of time-series data containing the historical records you wish to forecast and an optional secondary dataset of all relevant item attributes, you can generate a model that makes real-time predictions, saving time, money, and resources.
Dataset and Feature Group Requirements
This section specifies the Datasets / Feature Groups requirements to successfully train a Real-Time Forecasting model. Feature requirements include recommendations on additional datasets that might enhance model performance.
Training Models - Training Options and Metrics
This section describes all the available model training options that can be used to create a Real-Time Forecasting model. You can utilize the metric explanations to better understand how they measure the performance of the model you trained.
Evaluating Predictions
This section contains a quick model evaluation guide that helps you understand how well your model is performing.
Prediction API
This section discusses the prediction API method so that you could properly generate predictions from the model you deployed.