Decision Tree (Classification)

🧩 Problem Statement and Goal of Project

The goal of this project is to perform classification using the Decision Tree algorithm on a dataset loaded from a CSV file (Social_Network_Ads.csv). The classification task involves predicting whether a user purchases a product based on features such as age and estimated salary.

This project applies a Decision Tree Classifier to predict whether a user will purchase a product based on social network ad data. The dataset contains user attributes such as age and estimated salary, and the binary target variable indicates whether the product was purchased (1) or not (0).

The goal is to:

• Understand how Decision Trees partition feature space.
• Train and evaluate a model on real-world user behavior data.
• Visualize decision boundaries to explain classification.

🛠 Solution Approach

The notebook follows a well-structured classification pipeline:

1. Data Loading:

   • Read from Social_Network_Ads.csv

2. Preprocessing:

   • Extract features and labels.
   • Apply StandardScaler for feature normalization.

3. Train-Test Split:

   • 75% training / 25% test split using train_test_split.

4. Model Training:

   • Use DecisionTreeClassifier with 'entropy' criterion.

5. Prediction:

   • Test prediction for a sample input ([30, 87000])

6. (Optional): Visualizations and interpretability analysis.

🧰 Technologies & Libraries Used

• Python 3
• pandas, numpy – data manipulation
• scikit-learn – model training, preprocessing, evaluation
• matplotlib, seaborn – plotting and visualization

🚀 Installation & Execution Guide

Installation: Not explicitly provided, but typically requires:

pip install numpy pandas matplotlib scikit-learn

Execution:

1. Ensure Social_Network_Ads.csv is present in the working directory.
2. Run the notebook Decision Tree (Classification).ipynb sequentially.

📊 Key Results / Performance

• Confusion matrix used to evaluate performance.
• No specific accuracy metrics (e.g., precision, recall, F1) were calculated.
• Visualizations suggest decent separation of classes by the decision tree.
• Successfully trained a Decision Tree model using entropy criterion.
• Sample test prediction performed for [Age=30, Salary=87000].
• Feature scaling significantly improved decision boundary sharpness.

🖼️ Screenshots / Sample Outputs

• Plots included:

  • Decision boundaries for training set
  • Decision boundaries for test set

📚 Additional Learnings / Reflections

• Learned how decision trees split based on information gain (entropy).
• Emphasized the role of feature scaling in decision-based models.
• Understood basic interpretability of Decision Trees in classification tasks.

📁 Dataset Description

• File: Social_Network_Ads.csv

• Features:

  • Age: Numerical
  • EstimatedSalary: Numerical

• Target:

  • Purchased: Binary label (0 or 1)

• Goal: Predict purchase behavior from demographic features.

---

⚙️ Installation & Execution Guide

1. Clone the repository or download the .ipynb file.

2. Install dependencies:

   pip install pandas numpy scikit-learn matplotlib seaborn

3. Make sure Social_Network_Ads.csv is in the same directory.

4. Launch the notebook:

   jupyter notebook "Decision Tree (Classification).ipynb"

👤 Author

mehran Asgari

Email: [email protected] GitHub: https://github.com/imehranasgari

---

📄 License

This project is licensed under the Apache 2.0 License – see the LICENSE file for details.

---

💡 Some interactive outputs (e.g., plots, widgets) may not display correctly on GitHub. If so, please view this notebook via nbviewer.org for full rendering.