Machine Learning Classification Pipeline

A Python-based end-to-end machine learning project that demonstrates classification pipelines using scikit-learn, XGBoost, and TensorFlow. This repository includes implementations for training Random Forest classifiers on fertilizer recommendation datasets with complete data preprocessing, model training, and evaluation workflows.

📋 Table of Contents

• Overview
• Features
• Project Structure
• Stack
• Installation
• Usage
• Pipeline Architecture
• Results

📖 Overview

This project demonstrates best practices for building scalable ML pipelines with scikit-learn. It includes:

• Data acquisition from Google Drive
• Automated data preprocessing and cleaning
• Custom transformer implementations
• Model training with hyperparameter tuning
• Comprehensive evaluation metrics

The project focuses on a fertilizer recommendation classification task using custom transformer classes that integrate seamlessly with scikit-learn's pipeline architecture.

✨ Features

• Google Drive Integration: Direct data loading from Google Drive using OAuth2 or public URLs
• Custom Transformers: Reusable, modular preprocessing components (NullChecker, LabelEncoderTransformer, StandardScalerTransformer)
• Multiple ML Algorithms: RandomForest and XGBoost implementations
• Complete Pipeline: End-to-end workflow from data ingestion to model evaluation
• Flexible Authentication: Two approaches—OAuth2 flow and direct URL access

📁 Project Structure

.
├── RandomForestClassifier.py      # Main pipeline with Google Drive OAuth2 authentication
├── RandomForestClassifier2.py     # Simplified variant using direct Google Drive URL
├── requirements.txt               # Python dependencies
├── input.txt                      # Configuration file (stores Google Drive file_id)
├── .devcontainer/                 # Dev container setup for consistent environments
└── README.md                       # This file

🛠 Stack

Component	Technology
Language	Python 3.7+
ML Frameworks	scikit-learn 1.3.1, XGBoost 1.7.6, TensorFlow
Data Processing	pandas, NumPy
Visualization	matplotlib, seaborn 0.12.2
Cloud Integration	Google Drive API (googleapiclient 2.98.0)
Authentication	google-auth 2.23.0, google-auth-oauthlib 1.0.0

📦 Installation

Prerequisites

• Python 3.7 or higher
• pip package manager
• Google account (for accessing datasets on Google Drive)

Setup

1. Clone the repository:

git clone https://github.com/ssampathkumar104/machinelearning.git
cd machinelearning

2. Install dependencies:

pip install -r requirements.txt

3. Set up Google Drive credentials (for RandomForestClassifier.py):
   • Download credentials.json from Google Cloud Console
   • Place it in the project root directory
   • Create input.txt with your Google Drive file ID:

     file_id=YOUR_DRIVE_FILE_ID
     

🚀 Usage

Option 1: Using OAuth2 Authentication (RandomForestClassifier.py)

python RandomForestClassifier.py

This script:

• Authenticates with Google Drive using OAuth2
• Downloads the CSV file specified in input.txt
• Displays data overview (columns, shape, missing values, duplicates)
• Trains a Random Forest classifier with custom preprocessing steps
• Outputs accuracy and classification report

Option 2: Using Direct Google Drive URL (RandomForestClassifier2.py)

python RandomForestClassifier2.py

This simplified version:

• Uses a hardcoded public Google Drive file URL (no authentication needed)
• Executes the same preprocessing and training pipeline
• Provides the same evaluation metrics

🔄 Pipeline Architecture

Both implementations follow this modular pipeline structure:

Raw Data (CSV from Google Drive)
    ↓
[NullChecker] → Handle missing values (fill with 0)
    ↓
[LabelEncoderTransformer] → Encode categorical variables
    ↓
[StandardScalerTransformer] → Normalize numeric features
    ↓
[DataSplitter] → Split into train/test sets (75/25)
    ↓
[ModelTrainer] → Train RandomForest/XGBoost classifier
    ↓
[ModelSaver] → Persist model predictions
    ↓
Evaluation Metrics (Accuracy, Classification Report)

Custom Transformers

• NullChecker: Fills null values with 0
• LabelEncoderTransformer: Converts categorical features to numeric labels
• StandardScalerTransformer: Standardizes numeric features to mean=0, std=1
• DataSplitter: Splits data into train/test sets (configurable ratio)
• ModelTrainer: Trains ensemble models with configurable hyperparameters
• ModelSaver: Persists model predictions to pickle files

📊 Results

The model outputs:

• Accuracy Score: Overall classification accuracy on the test set
• Classification Report: Precision, recall, and F1-score per class
• Saved Model: Predictions saved to final_rf_model.pkl

Example output:

Accuracy: 0.92
Classification Report:
              precision    recall  f1-score   support
           0       0.91      0.93      0.92       245
           1       0.93      0.91      0.92       255

🔧 Configuration

Edit pipeline parameters in the ModelTrainer initialization:

ModelTrainer(
    n_estimators=50,        # Number of trees in the forest
    max_depth=10,           # Maximum tree depth
    min_samples_split=2,    # Minimum samples to split a node
    min_samples_leaf=1,     # Minimum samples in leaf node
    criterion='gini',       # Split quality criterion
    bootstrap=True          # Use bootstrap samples
)

📝 Notes

• Both scripts target the fertilizer recommendation classification task (target_feature='fertilizer_name')
• The project is designed for educational purposes and demonstrates ML best practices
• Custom transformers are compatible with scikit-learn's Pipeline and GridSearchCV
• TensorFlow is listed in requirements but not currently used; reserved for future deep learning implementations

🤝 Contributing

Contributions welcome! Feel free to:

• Improve model performance
• Add additional preprocessing techniques
• Extend to other ML algorithms
• Enhance documentation

📄 License

This project is open source and available for educational use.

---

Last Updated: November 2024