Decision Tree for WiFi Room Classification

A complete decision tree implementation using the ID3 algorithm to classify rooms based on WiFi signal strengths. Includes visualization, cross-validation, and comprehensive evaluation metrics.

Features

• ID3 Decision Tree: Information gain-based splitting with entropy calculation
• Smart Visualization:
  • Color-coded nodes (blue=decision, green=leaf)
  • Automatic layout preventing overlap
  • Auto-sizing based on tree dimensions
  • High-resolution PNG export
• Evaluation: 10-fold cross-validation with confusion matrix, precision, recall, and F1-scores
• Prediction: Single sample and batch prediction support

Quick Start

# Activate virtual environment
source venv/bin/activate  # Windows: venv\Scripts\activate

# Run complete evaluation
python main.py

This runs the full pipeline (tree building, visualization, cross-validation) in ~30 seconds without user interaction.

Installation

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Requirements: Python 3.12+, NumPy 2.1.1, Matplotlib 3.9.2

Usage

Complete Evaluation (Recommended)

python main.py

What it does:

1. Builds decision trees on clean and noisy datasets
2. Saves visualizations to clean_tree_visualization.png and noisy_tree_visualization.png
3. Runs 10-fold cross-validation on both datasets
4. Displays confusion matrices and per-class metrics

Expected Results:

• Clean dataset: ~97.4% accuracy, depth ~14, ~44 leaves
• Noisy dataset: ~80.4% accuracy, depth ~18, ~335 leaves

Visualization Only

python visualize_demo.py

Creates tree_visualization.png with a tree trained on the clean dataset.

Using in Your Code

Build and Visualize a Tree

from decision_tree_builder import decision_tree_learning
import numpy as np

# Load dataset (features + label in last column)
dataset = np.loadtxt('wifi_db/clean_dataset.txt')

# Build tree
tree = decision_tree_learning(dataset, depth=0)

# Visualize (automatic sizing, non-blocking)
tree.visualize_tree(save_path='my_tree.png', show=False)

# Or with custom size and display
tree.visualize_tree(figsize=(20, 12), save_path='my_tree.png', show=True)

# Make predictions
prediction = tree.predict(sample_features)
batch_predictions = tree.predict_batch(multiple_samples)

# Get metrics
print(f"Depth: {tree.get_depth()}, Leaves: {tree.get_leaf_count()}")

Run Cross-Validation

from evaluation import k_fold_cross_validation

# Run 10-fold CV
results = k_fold_cross_validation(dataset, k=10, random_seed=42)

# Access results
print(f"Average accuracy: {results['average_accuracy']:.3f}")
print(f"Confusion matrix:\n{results['confusion_matrix']}")

# Per-class metrics
for label, precision, recall, f1 in zip(
    results['labels'], results['precision'], 
    results['recall'], results['f1']
):
    print(f"Room {label}: P={precision:.3f}, R={recall:.3f}, F1={f1:.3f}")

Single Train/Test Split

from evaluation import evaluate

# Split data
train_data = dataset[:1600]
test_data = dataset[1600:]

# Train and evaluate
tree = decision_tree_learning(train_data, depth=0)
accuracy = evaluate(test_data, tree)
print(f"Accuracy: {accuracy:.3f}")

Visualization Options

Parameters

tree.visualize_tree(
    figsize=None,        # (width, height) or None for auto-sizing
    save_path=None,      # Path to save PNG, or None for display only
    show=True            # True to display, False to save without blocking
)

Examples

# Automatic sizing, save only (recommended for automation)
tree.visualize_tree(save_path='tree.png', show=False)

# Custom size with display
tree.visualize_tree(figsize=(20, 12), save_path='tree.png', show=True)

# Display only
tree.visualize_tree(show=True)

Project Structure

.
├── decision_tree_builder.py    # ID3 algorithm implementation
├── tree_node.py                # TreeNode class with visualization
├── evaluation.py               # Cross-validation and metrics
├── main.py                     # Complete evaluation pipeline
├── visualize_demo.py           # Visualization demo
├── requirements.txt            # Dependencies
├── wifi_db/
│   ├── clean_dataset.txt      # Clean WiFi signals (2000×8)
│   └── noisy_dataset.txt      # Noisy WiFi signals (2000×8)
└── Output:
    ├── clean_tree_visualization.png
    └── noisy_tree_visualization.png

Dataset Format

• 7 feature columns: WiFi signal strengths (dBm) from AP1-AP7
• 1 label column: Room number (1, 2, 3, or 4)
• 2000 samples: 500 per room class

Troubleshooting

Virtual environment not activated:

source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

Module not found:

pip install -r requirements.txt

Visualization doesn't display:

• By default, main.py saves without displaying (show=False)
• To display: tree.visualize_tree(save_path='tree.png', show=True)

Nodes overlapping:

• Use automatic sizing: tree.visualize_tree(save_path='tree.png')
• Or increase size: tree.visualize_tree(figsize=(30, 20), save_path='tree.png')

Algorithm Details

ID3 Decision Tree:

• Entropy-based information gain for split selection
• Binary splits on continuous features (threshold-based)
• Stops when: all samples same class, no features remain, or dataset empty

10-Fold Cross-Validation:

• Dataset shuffled and split into 10 equal folds
• Each fold used once as test set
• Fixed random seed (42) for reproducibility

Performance

• Training: ~1-2 seconds per tree
• Cross-validation: ~20-30 seconds per dataset
• Visualization: ~5-10 seconds
• Memory: ~50-100 MB

License

Coursework project for COMP60012: Introduction to Machine Learning at Imperial College London.