Using confusion matrix to evaluate the performance of decision tree

Question

Hi I am trying to use Confusion Matrix to evaluate the performance of decision tree. I have written the function for it but not sure how to use the predicted labels and test labels(not sure which data frame to create or code). Please can someone look and suggest what can be done here(look at the last lines of code)?

Code:

#Importing Pandas library
import pandas as pd
import numpy as np

#Reading file into data frame
csv_path = 'winequality-white-v3.csv'
df0 = pd.read_csv(csv_path) 

#Filtering missing values
missingchlorides = pd.isna(df0["chlorides"])
missingIndices = df0[missingchlorides].index

#Replacing missing values by mean
meanchlorides = float(df0["chlorides"].mean())
df0["chlorides"].where(~ missingchlorides, meanchlorides, inplace=True)

#Deleteing missing values at the EOF
missing_winequailty = pd.isna(df0["alcohol"])
missingIndices = df0[missing_winequailty].index
df1 = df0.drop(missingIndices, axis=0)

#Transformaing the quality variable to 0 and 1
def transformQuality(quality):
    if quality == 5:
        quality = 0
    else:
        quality = 1
    return quality

df1["quality"] = df1["quality"].apply(transformQuality)

#Saving into a csv file
df1.to_csv('filtered-winequality-white-v3.csv') 

*********************************** HERE ****************************
#PART i:
    
#Importing Orange Library
#Importing “SklTreeLearner”, cross validation, scoring and confusion matrix
from Orange.data import Table, Domain
from Orange.classification import SklTreeLearner
from Orange.evaluation import CrossValidation, scoring
from sklearn.metrics import confusion_matrix

#Reading the fileterd data
Filtered_data = Table.from_file('filtered-winequality-white-v3.csv')

#Defining features
feature_vars = list(Filtered_data.domain.variables[1:6])
class_label_var = Filtered_data.domain.variables[7]

#Defining domain
winequality_domain = Domain(feature_vars, class_label_var)
Filtered_data= Table.from_table(domain=winequality_domain, source=Filtered_data)

#Shuffling and splitting data for traning and testing
Filtered_data.shuffle()
train_data_tab = Filtered_data[:1800]
test_data_tab = Filtered_data[1800:]

#Creating tree learner and decision tree
tree_learner = SklTreeLearner()
decision_tree = tree_learner(train_data_tab)

#Creating external prediction function having the decision tree and the input  
def decision_tree_predict(d_tree, input_data):
    predicted_label_vals = d_tree(input_data)

    predicted_labels = []
    for val in predicted_label_vals:
        predicted_labels.append(input_data.domain.class_var.values[int(val)])

    return predicted_labels


#PART ii:
    
#Estimating the accuracy of decision_tree using the entire testing data
#Performance estimation by comparing the predicted class with the actual class 
predicted_class_labels = decision_tree_predict(decision_tree, test_data_tab[:])

num_of_test_samples = len(predicted_class_labels)
num_of_correct_predictions = 0

for i in range(num_of_test_samples):
    if predicted_class_labels[i] == test_data_tab[i, test_data_tab.domain.class_var]:
        num_of_correct_predictions += 1

accuracy = num_of_correct_predictions / num_of_test_samples
print('Accuracy = {:.3f}'.format(accuracy))

#Evaluating the performance of a decision tree classifier using cross-validation
eval_results = CrossValidation(Filtered_data, [tree_learner], k=10)
print("Accuracy: {:.3f}".format(scoring.CA(eval_results)[0]))
print("AUC: {:.3f}".format(scoring.AUC(eval_results)[0]))



'''
print(test_data_tab[:,5:])
print(predicted_class_labels)


y_true = test_data_tab[:,5:]
y_pred = predicted_class_labels
confusion_matrix(y_true, y_pred, labels= ['0','1'])
'''

Thanks!

Answer 1

Here is a code for the tree and confusion matrix:

# Create the tree
tree = DecisionTreeClassifier(max_depth=6, class_weight='balanced')
tree.fit(X_train,y_train)

#create array of probabilities
y_test_predict_proba = tree1.predict_proba(X_test)

# calc confusion matrix
y_test_predict = tree.predict(X_test[columns])
print("Confusion Matrix Tree : \n", confusion_matrix(y_test, y_test_predict),"\n")
print("The precision for Tree is ",precision_score(y_test, y_test_predict)) 
print("The recall for Tree is ",recall_score(y_test, y_test_predict),"\n")  

You did not post your tree code, so I could not have applied my code on your tree