Customer Segmentation with Machine Learning & CodeIgniter 4

End-to-end customer segmentation system using Machine Learning (K-Means clustering) integrated with CodeIgniter 4 web framework.

Features

• Machine Learning Pipeline: Complete ML pipeline with preprocessing, feature engineering, and K-Means clustering
• Model Persistence: Save and load trained models for predictions
• Web Dashboard: Interactive dashboard showing cluster statistics and distributions
• Real-time Predictions: Predict customer segments for new data
• Visualization: 2D PCA visualization of customer clusters
• Batch Processing: Upload CSV files for batch predictions
• RESTful API: JSON API endpoints for integration

Architecture

MachineLearningwithCodeIgniter/
├── ml_pipeline/              # Machine Learning Pipeline
│   ├── customer_segmentation.py   # Main training pipeline
│   ├── predict_api.py             # Prediction API
│   ├── train_model.sh             # Training script
│   └── requirements.txt           # Python dependencies
├── models/                   # Trained models & data
│   ├── segmentation_model.pkl     # Trained model
│   ├── cluster_profiles.json      # Cluster statistics
│   └── segmented_customers.csv    # Segmented data
├── app/
│   ├── Controllers/
│   │   └── Segmentation.php       # Main controller
│   ├── Models/
│   │   └── CustomerModel.php      # Data model
│   ├── Views/
│   │   └── segmentation/          # Web interface
│   └── Config/
│       └── Routes.php             # Routes configuration
├── public/                   # Web root
└── bank.csv                  # Source data

Requirements

Backend (PHP)

• PHP 8.1 or higher
• CodeIgniter 4.x
• Composer

Machine Learning (Python)

• Python 3.8+
• pandas
• numpy
• scikit-learn
• matplotlib
• seaborn

Installation

1. Install PHP Dependencies

composer install

2. Install Python Dependencies

cd ml_pipeline
pip3 install -r requirements.txt

3. Configure Environment

cp .env.example .env

Edit .env and configure your settings.

4. Train the Model

cd ml_pipeline
bash train_model.sh

This will:

• Load the bank.csv data
• Preprocess features
• Train K-Means clustering model
• Generate cluster profiles
• Save model and segmented data

Usage

Start the Web Server

php spark serve

Visit: http://localhost:8080

Web Interface

1. Dashboard (/segmentation)

   • View cluster statistics
   • See cluster distribution
   • Train/retrain model

2. Predict (/segmentation/predict)

   • Input customer data
   • Get real-time segment prediction
   • View cluster profile

3. Visualize (/segmentation/visualize)

   • 2D scatter plot of clusters
   • PCA visualization
   • Interactive charts

4. Customers (/segmentation/customers)

   • Browse segmented customers
   • Export data

API Endpoints

Get Cluster Statistics

GET /segmentation/getClusterStats

Response:

{
  "success": true,
  "profiles": {
    "0": {
      "cluster_id": 0,
      "size": 2500,
      "percentage": 25.0,
      "features": {...}
    }
  },
  "distribution": [...]
}

Predict Single Customer

POST /segmentation/predictSegment
Content-Type: application/json

{
  "age": 35,
  "job": "management",
  "marital": "married",
  "education": "tertiary",
  "balance": 5000,
  ...
}

Response:

{
  "success": true,
  "prediction": {
    "cluster": 2,
    "cluster_profile": {...},
    "confidence": 0.85
  }
}

Get Visualization Data

GET /segmentation/getVisualizationData

Train Model

POST /segmentation/trainModel

Python API (CLI)

cd ml_pipeline

# Predict single customer
python3 predict_api.py '{"age": 35, "job": "management", "balance": 5000, ...}'

# Batch prediction
python3 predict_api.py '[{...}, {...}, ...]'

Machine Learning Pipeline

Features Used

• Demographics: age, job, marital, education
• Financial: balance, housing, loan, default
• Campaign: contact, day, month, duration, campaign, pdays, previous, poutcome
• Target: deposit

Preprocessing Steps

1. Label encoding for categorical variables
2. Feature scaling using StandardScaler
3. PCA for dimensionality reduction (visualization)

Clustering

• Algorithm: K-Means
• Number of clusters: 4 (configurable)
• Distance metric: Euclidean

Model Outputs

• Cluster assignments: 0, 1, 2, 3
• Cluster profiles: Statistical summaries for each segment
• PCA coordinates: For visualization

Data

Source

• Dataset: Bank Marketing Dataset (bank.csv)
• Records: 11,163 customers
• Features: 17 attributes

Columns

• age: Customer age
• job: Type of job
• marital: Marital status
• education: Education level
• default: Has credit in default?
• balance: Average yearly balance (euros)
• housing: Has housing loan?
• loan: Has personal loan?
• contact: Contact communication type
• day: Last contact day of month
• month: Last contact month
• duration: Last contact duration (seconds)
• campaign: Number of contacts during campaign
• pdays: Days since last contact
• previous: Number of contacts before campaign
• poutcome: Outcome of previous campaign
• deposit: Has term deposit?

Customization

Change Number of Clusters

Edit ml_pipeline/customer_segmentation.py:

pipeline = CustomerSegmentationPipeline(n_clusters=5)  # Change from 4 to 5

Add New Features

1. Update feature selection in preprocess_data() method
2. Retrain model
3. Update prediction form in views

Modify Clustering Algorithm

Replace K-Means with other algorithms:

from sklearn.cluster import DBSCAN, AgglomerativeClustering

# Instead of KMeans
self.model = DBSCAN(eps=0.5, min_samples=5)

Deployment

Production Checklist

• Set CI_ENVIRONMENT = production in .env
• Configure production database
• Set proper file permissions
• Configure web server (Apache/Nginx)
• Install Python dependencies on server
• Train model with production data
• Enable HTTPS
• Configure CORS if needed
• Set up monitoring and logging

Web Server Configuration

Apache (.htaccess):

RewriteEngine On
RewriteCond %{REQUEST_FILENAME} !-f
RewriteCond %{REQUEST_FILENAME} !-d
RewriteRule ^(.*)$ index.php/$1 [L]

Nginx:

location / {
    try_files $uri $uri/ /index.php?$query_string;
}

Troubleshooting

Model Not Found Error

• Train the model first: bash ml_pipeline/train_model.sh
• Check models/ directory exists and has permissions

Python Not Found

• Check Python installation: python3 --version
• Update path in controller: $this->pythonPath = 'python3'

Permission Errors

chmod +x ml_pipeline/train_model.sh
chmod -R 755 models/

Prediction Fails

• Ensure all required fields are provided
• Check input data format matches training data
• Verify model is loaded correctly

Performance

• Training time: ~30 seconds (11K records)
• Prediction time: <100ms per customer
• Model size: ~500KB
• Memory usage: ~50MB

Future Enhancements

• Add more clustering algorithms
• Implement cluster comparison
• Add customer lifetime value prediction
• Create automated retraining schedule
• Add A/B testing framework
• Implement real-time streaming predictions
• Add model versioning
• Create admin panel for model management

License

MIT License

Credits

Developed using:

• CodeIgniter 4
• Scikit-learn
• Bootstrap 5
• Chart.js

---

Note: This is a demonstration project for educational purposes. For production use, add proper authentication, authorization, input validation, and error handling.