Reading the Correlation Graph in Driverless AI
- Predictive AI
- H2O Driverless AI
This lesson covers the correlation graph in Driverless AI's visualizations page, an interactive 2D map showing all 26 features in a dataset and the strength of correlation between them as colored connecting lines. It explains the color scale, from violet for low correlation to red for high, and shows how dragging the correlation threshold slider higher filters out weaker connections in real time, narrowing the view to only the most highly correlated features. It also notes the four features in this dataset with no lines at all, meaning no correlation to anything else, and points to the graph's SVG export and help options.
Key takeaways
- How to read the correlation graph's color scale, from violet edges for low correlation to red edges for high correlation
- How dragging the threshold slider higher filters out weaker connections and narrows the graph to the strongest correlations
- Why four features in this dataset, including ID and Age, show no connecting lines at all on the correlation graph
- What raising the threshold to 0.99 revealed here: only monthly income and annual income stayed connected as highly correlated
- How to export the correlation graph as an SVG file or reset its threshold back to the default view
Start owning your AI