The BigML models layout is very good at preventing the decision tree model from becoming too wide, while enabling each branch to be clearly displayed. However, when the model is still too wide to be displayed on a normal browser BigML only shows the parts of the model that have at least 1% support (i.e., 1% of the dataset instances). The width of the branches also represent the amount of data used in each branch and nodes, the wider the branch, the more data instances being analyzed in that branch. Therefore, the lines at the top of the model tend to be wider than the ones at the bottom. In this blog post you can learn how to filter branches by SUPPORT or CONFIDENCE.
Along the side of the model there is a prediction path showing the information and fields (making up a set of rules) used to make that branching decision, which leads to a given prediction.
The colors of the model represent the different fields the branch uses. As you can see in the above example, the color of the fields on the prediction path matches the color of the nodes in the model. This makes it easier to identify important fields that commonly occur all throughout the model. Notice that whenever you mouse over a branch in the model (shown by a thicker, darker line), the corresponding path information shows up on the right hand side.
Finally, BigML uses green color for the prediction nodes, which are the terminal nodes of each path and they give you the final prediction.
Would you like to know how BigML represents the decision tree models in the Sunburst view? Click here to discover a different way to visualize the model above.