Support on Root Cause analysis

[NKMatha]

Hi team,

Am not clear with decision trees generated for Root Cause analysis. Can some please explain what does it representing for my data.

Thnaks.

[fit-alessandro-berti]

For the "decision tree" visualization of Scikit-Learn, in general you should interpret it as follows:

1. Nodes: These are the "decision" points where the data is split.
2. Root Node: This is the topmost node of the tree where the first split is made.
3. Internal Node: These are the nodes where subsequent splits happen.
4. Leaf Node/Terminal Node: Nodes at the bottom of the tree where a prediction is made, with no further splitting.
5. Edges or Branches: These are the lines that connect nodes, representing the flow from one question to the next.
6. Splitting Criteria: This is the condition based on which the split has been made at a particular node. It usually involves a feature and a threshold value.
7. Gini Impurity: A measure of purity used at the node. The Gini Impurity ranges from 0 (all elements at the node belong to the same class) to 0.5 (elements are randomly distributed across various classes). In case of a binary classification, it's 0.5, for multiple classes, it can go higher.
8. Samples: The number of observations in the dataset that reach that node.
9. Value: The class distribution of the samples at that node. For example, if you are predicting three different classes, the value might look like [34, 10, 5], which means there are 34 samples of class 1, 10 samples of class 2, and 5 samples of class 3.
10. Class: The class that is the most prevalent in that node. In a leaf node, this represents the predicted class for samples that end up in that leaf.

[fit-alessandro-berti]

The case-level features of pm4py are generally computed using one-hot encoding for categorical values, and reporting the last numeric value of numeric features.

[NKMatha]

Thanks @fit-alessandro-berti