Great! This second notebook is a classification lab using Decision Trees on a medical dataset called drug200.csv, which contains patient information and the type of drug prescribed to them.

---

🔍 Detailed Explanation of the Notebook

---

✅ Objective

To build a Decision Tree Classifier that predicts the drug to prescribe to a patient based on features like age, sex, blood pressure, cholesterol, and sodium/potassium levels.

---

📦 1. Package Installation

!pip install numpy==2.2.0
!pip install pandas==2.2.3
!pip install scikit-learn==1.6.0
!pip install matplotlib==3.9.3

These ensure compatible versions of important libraries are used:

• NumPy, Pandas: Data manipulation
• Matplotlib: Plotting
• scikit-learn: Machine learning models

---

📥 2. Import Libraries

import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier, plot_tree
from sklearn import metrics

It imports tools for:

• Preprocessing: Label encoding categorical variables
• Modeling: Training a decision tree
• Evaluation: Accuracy, confusion matrix, etc.
• Visualization: Drawing the tree

---

📊 3. Load Dataset

my_data = pd.read_csv('drug200.csv')

Dataset contains:

• Features: Age, Sex, Blood Pressure, Cholesterol, Na_to_K (sodium to potassium ratio)
• Target: Drug (DrugA, DrugB, ..., DrugY)

---

🔍 4. Data Inspection

my_data.info()

Gives insight into:

• Number of records
• Data types
• Non-null values

---

🔧 5. Preprocessing

label_encoder = LabelEncoder()
my_data['Sex'] = label_encoder.fit_transform(my_data['Sex'])
my_data['BP'] = label_encoder.fit_transform(my_data['BP'])
my_data['Cholesterol'] = label_encoder.fit_transform(my_data['Cholesterol'])

Transforms categorical features into numbers so they can be used in the decision tree:

• Sex: Male/Female → 1/0
• BP: HIGH/NORMAL/LOW → e.g. 2/1/0
• Cholesterol: HIGH/NORMAL → e.g. 1/0

---

Here’s a continued and detailed breakdown of the notebook, step-by-step:

---

🧼 6. Missing Data Check

my_data.isnull().sum()

This checks for missing values in each column. All values are expected to be complete.

---

🔢 7. Convert Target to Numeric

custom_map = {'drugA': 0, 'drugB': 1, 'drugC': 2, 'drugX': 3, 'drugY': 4}
my_data['Drug_num'] = my_data['Drug'].map(custom_map)

• Converts the Drug column (target class) into numeric values.
• Makes it easier to work with classification models.

---

🧠 8. Dataset Info After Encoding

my_data.info()

Now the data types show encoded categorical columns and the numeric target column (Drug_num).

---

📉 9. Histograms of All Features

my_data.hist(bins=30, color='r', figsize=(16, 16))

This visualizes distributions of numeric features like Age and Na_to_K, which helps understand their spread.

---

📈 10. Feature Correlation

data = my_data.select_dtypes(include='number')
corr = data.corr()

This selects only numerical columns to compute the correlation matrix, which shows how features are related to each other.

---

🔥 11. Heatmap of Correlations

import seaborn as sns
plt.figure(figsize=(8, 8))
sns.heatmap(corr, annot=True)
plt.show()

• Visual tool showing how strongly features are correlated.
• Helpful for identifying redundant or predictive features.

---

📊 12. Drug Class Distribution (Twice)

drugs = my_data['Drug'].value_counts()
plt.bar(drugs.index, drugs.values, color='r')

and

category_counts = my_data['Drug'].value_counts()
plt.bar(category_counts.index, category_counts.values, color='blue')

• These two cells plot how often each drug label occurs in the dataset.
• Confirms class balance or imbalance (important for training accuracy).

---

Now let’s dive into the core modeling part of the notebook: building and evaluating the Decision Tree classifier.

---

🌲 13. Define Features & Target

y = my_data['Drug']
X = my_data.drop(['Drug', 'Drug_num'], axis=1)

• X: All features like Age, Sex, BP, Cholesterol, Na_to_K
• y: The actual drug class (string labels)

---

🧪 14. Split Dataset

X_trainset, X_testset, y_trainset, y_testset = train_test_split(X, y, test_size=0.3, random_state=32)

• 70% for training, 30% for testing
• random_state=32 ensures reproducibility

---

🧠 15. Train a Decision Tree

drugTree = DecisionTreeClassifier(criterion="entropy", max_depth=4)
drugTree.fit(X_trainset, y_trainset)

• Uses Entropy as the splitting criterion (information gain)
• max_depth=4 limits the depth to prevent overfitting

---

📊 16. Predict on Test Set

tree_predictions = drugTree.predict(X_testset)

Makes predictions for unseen data (testing set).

---

🎯 17. Accuracy Evaluation

print("Decision Trees's Accuracy: ", metrics.accuracy_score(y_testset, tree_predictions))

Compares predicted vs. actual drug labels and prints the model's accuracy.

---

🌳 18. Visualize the Tree

plot_tree(drugTree)
plt.show()

• Displays the actual decision tree, showing:
  • Feature splits
  • Threshold values
  • Class outcomes

---

🧪 19. Another Tree with Different Depth

tree_model = DecisionTreeClassifier(criterion='entropy', max_depth=3)
tree_model.fit(X_trainset, y_trainset)

• A second tree is built with max_depth=3 to compare model behavior with a shallower tree.

---

✅ 20. Accuracy of Second Tree

pred = tree_model.predict(X_testset)
print(f"Accuracy = {np.round(100*metrics.accuracy_score(y_testset, pred), 2)}%")

• Calculates accuracy for the second tree
• Displays in percentage format

---

🌿 21. Visualize Second Tree

plot_tree(tree_model)
plt.show()

• Shows the structure of the second (shallower) tree.

---

✅ Summary of Model Section:

Step	Description
Model	Decision Tree Classifier
Criterion	Entropy (Information Gain)
Output	Drug prescription (DrugA–Y)
Accuracy	Evaluated on test set
Visualization	Tree structure shown using plot_tree()

---