Interactive Graph

Hopsworks provides an interactive graph to help you explore the statistics computed on your feature data more efficiently and help you identify anomalies faster. The graph lives on the Feature Monitoring tab of a Feature Group or Feature View, one page per feature monitoring configuration.

Select a feature monitoring configuration#

First, you need to select a feature monitoring configuration to visualize. The dropdown in the page header lists every configuration defined on the Feature Group or Feature View.

Select a statistics metric to visualize#

Below the graph, the feature table has one checkbox per feature and statistics metric. Ticking a checkbox plots that metric over time.

Visualize multiple metrics simultaneously#

Several metrics can be visualized at the same time on the graph. Tick more than one checkbox in the feature table, and each metric gets its own colour and legend entry.

Show reference statistics#

In feature monitoring configurations with reference windows, you can also visualize the reference values by enabling Reference under Show above the graph. Reference values are drawn as a dashed line: statistics computed over time, or a horizontal line for a specific value.

Show threshold bounds#

In addition to reference windows, you can define thresholds to automate the identification of data points as anomalous values. A threshold can be absolute, or relative to the statistics values under comparison. You can visualize the threshold bounds as a band around the reference line by enabling Threshold under View.

Highlight shifted data points#

If a reference window and threshold are provided, data points that fall out of the threshold bounds are considered anomalous values. You can highlight these data points by enabling Shift detected under Show. The feature table below the graph flags the same features in its Shift column.

Visualize the computed differences between statistics#

Alternatively, you can change the time series to show the differences computed between detection and reference statistics rather than the statistics values themselves. You can achieve that by enabling Difference under View. The threshold then shows as a horizontal line.

Configuration summary and controls#

The card at the top of a configuration page summarizes the detection and reference windows, the statistics comparison criteria and the job schedule. From there you can trigger the statistics comparison manually with Run once, or pause the schedule of the feature monitoring job with Disable.

List of configurations#

The Feature Monitoring tab itself lists all feature monitoring configurations defined for the Feature Group or Feature View, with their status and next scheduled check.