Decision Tree Root Cause Analysis

decision_tree_root_cause.py

import graphviz
import pandas as pd
import numpy as np
import sklearn.impute
import sklearn.tree

# Load the data from disk
df_good = pd.read_csv("~/Downloads/good_data_sample.csv")
df_bad = pd.read_csv("~/Downloads/bad_data_sample.csv")

# Create a binary outcome response
df = pd.concat([df_good, df_bad])

# Create a feature matrix
X = pd.get_dummies(df)

# Create a response variable
zeros = np.repeat(0, len(df_good))
ones = np.repeat(1, len(df_bad))
Y = np.concatenate([zeros, ones])

# Impute missing feature values with their mean
imp = sklearn.impute.SimpleImputer()
imp.fit(X)
X_imputed = imp.transform(X)

# Fit a decision tree
random_state = 2
model = sklearn.tree.DecisionTreeClassifier(
    random_state=random_state, min_samples_leaf=100)
model.fit(X_imputed, Y)

# Visualize the decision tree
dot_data = sklearn.tree.export_graphviz(
    model, out_file=None, feature_names=list(X.columns),  
    class_names=["good data", "bad data"],
    filled=True, rounded=True)  
graphviz.Source(dot_data)