Figure 3.
The decision tree begins with the root node S E P 11 less than or equal to 1.0, with gini 0.453, samples 121, value 63, 119, and class Effect. Branches are labelled True on the left and False on the right. Each node contains a decision condition where applicable, followed by gini, samples, value, and class. Terminal nodes contain only gini, samples, value, and class. The left branch immediately ends in a terminal node with gini 0.0, samples 6, value 9, 0, and class Cause. The right branch continues to S E P 18 less than or equal to 2.5, then divides into a left terminal node with gini 0.28, samples 28, value 0, 47, class Effect, and a right branch leading to S E P 13 less than or equal to 1.0. The left branch of S E P 13 ends in a terminal node with gini 0.0, samples 8, value 10, 0, class Cause, while the right branch continues to S E P 15 less than or equal to 3.5. From this point the tree divides into two major branches. The left branch proceeds through S E P 12 less than or equal to 1.5, S E P 1 less than or equal to 2.5, S E P 14 less than or equal to 1.0, S E P 16 less than or equal to 1.0, S E P 3 less than or equal to 2.5, and S E P 10 less than or equal to 2.5, ending in multiple terminal Cause and Effect nodes with gini values ranging from 0.0 to 0.497. The right branch proceeds through S E P 9 less than or equal to 1.0, S E P 16 less than or equal to 3.5, S E P 8 less than or equal to 1.0, S E P 6 less than or equal to 2.5, S E P 19 less than or equal to 1.0, and S E P 17 less than or equal to 1.5, also terminating in multiple Cause and Effect nodes. Across the tree, decision nodes report gini values between 0.1 and 0.497, sample counts between 6 and 121, and value pairs corresponding to the two classes. Terminal nodes consistently have gini 0.0 except where further splits occur. The tree contains both Cause and Effect terminal classifications distributed throughout the left and right branches, with directional connecting lines linking every parent node to its child nodes.Random forest model visualization
Source(s): Authors’ own