NextGen experience > AI experimentation > Predictive experiments > Evaluate with model insights > Accuracy Over Space

Accuracy Over Space¶

Tab	Description
Performance	Helps to discover spatial patterns in prediction errors and visualize prediction errors on a map visualization. The tab provides a spatial residual mapping within an individual model and across all data partitions.

By default, Accuracy Over Space displays residual (prediction error) values on a unique map based on the validation partition. Co-located points are displayed using the average value of all points at that location. The map type and visualization settings can be adjusted to further fine-tune the display. The collapsible map legend updates as you change the settings. The following table describes the settings:

Setting	Description
Visualization	Sets the map type for the visualization.
Aggregation	Sets the arithmetic to use for co-located locations, either avg, min, max, count, or value. Count reports the count of each geometry in the dataset; for a hexagon or grid map of a geometry feature, it reports the sum of counts. Value displays the feature value (no aggregation) when all geometries have a count of 1.
Data selection	Sets which data partition to visualize, either validation, cross-validation, or holdout.
Metric type	Sets the value to report at each location, either: Actual: actual values from the dataset. Predicted: Values predicted by model. Residual: The difference between actual and predicted values.
Visualization settings	Allows you to set a variety of map viewing settings including opacity, coverage, elevation scale, percentile, radius, and intensity.

Map visualizations¶

The following map-type options are available from the Visualization dropdown.

Unique map¶

The Unique map shows individual unique points on the map. If points are on the same location, that information is reflected in the "Count" column of the tooltip shown when hovering over a point.

Kernel density map¶

A Kernel density map collects multiple observations within each given kernel and displays aggregated statistics with a color gradient. For location features, the count, min, max, and average can be selected from the Aggregation dropdown. For numeric features, the min, max, or average is available. For categorical features, the mode is displayed. Several visualization customizations are available in Visualization Settings.

Hexagon map¶

The Hexagon map is an aggregated view of points on the map, rolled up into hexagon bins. Select Hexagon map to display data as hexagon-shaped cells. When selecting Aggregation settings:

For location features, count, min, max, and average are available.
For numeric features, min, max, or average are available.
For categorical features, the mode is displayed.

Use the Visualization settings in the bottom-right of the map panel to adjust the settings.

Heatmap¶

The Heatmap layer shows the spatial distribution of data on the map, weighted by the selected metric. You can view heatmap visualizations for location and numeric features; it is not available for categorical features. Use the Visualization settings in the bottom-right of the map panel to adjust the settings.

Visualization settings¶

The following settings are available in the Visualization settings dropdown, with selection being dependent on map type. The following table describes each setting:

Setting	Description
Coverage	Sets the percentage of the radius that the hexagon displayed on the map occupies. The full cluster/hexagon is of width `Radius`. The hexagon displayed on the map occupies `Coverage %` at the center of the full hexagon.
Elevation scale	Controls the scale of how tall the clusters can be.
Intensity	Modifies the intensity of the color of each pixel; used by the heat map.
Opacity	Adjusts transparency of layers.
Percentile	As the slider moves, it removes a portion, by percentile, from the map. This setting is helpful for identifying outliers.
Radius	Controls the radius of the hexagon clusters; this is the precision of the aggregation.