🏨 Hotel Reservation Cancellation Prediction (MLOps)

This repository contains an end-to-end MLOps pipeline designed to predict whether a customer will honor or cancel their hotel reservation. The system leverages cloud data storage, robust model tracking, automated CI/CD pipelines, and serverless container deployment.

---

🏗️ System Architecture & Workflow

1. Data Layer: Raw reservation data is managed via automated ETL flows and stored securely in a Google Cloud Storage bucket.
2. Experimentation: Version control handles small tracking files while heavy assets are tracked via Git. Models are monitored across iterations using an MLflow tracking server.
3. Continuous Integration & Deployment: Commits to GitHub trigger automated Jenkins pipelines. Jenkins builds a Docker image via Docker-in-Docker (DinD), registers it to Google Container Registry (GCR), and ships it to Google Cloud Run.

---

📁 Repository Directory Structure

├── src/                      # Source code modules (Ingestion, Preprocessing, Training)
├── notebook/                 # Jupyter Notebooks for EDA and prototype testing
├── templates/                # HTML files for the Flask UI
├── static/                   # CSS and JavaScript assets
├── config/                   # Configuration files (config.yaml, model_params.yaml)
├── artifacts/                # Local data splits and serialized model outputs
├── pipeline/                 # Training and prediction orchestration scripts
├── utils/                    # Common helper utilities
├── Dockerfile                # Project container definition
├── requirements.txt          # Python dependencies
└── setup.py                  # Project package installation settings

---

⚙️ Local Development Setup

1. Environment Initialization

Isolate your development dependencies by initializing a clean virtual environment:

python -m venv venv

Activate the environment:

OS	Command
Windows (PowerShell)	venv\Scripts\activate
Linux / macOS	source venv/bin/activate

2. Dependency Installation

Install required libraries (including imbalanced-learn) and package the source directory in editable mode:

pip install -r requirements.txt
pip install -e .

---

☁️ Google Cloud Platform Configuration

1. Service Account Authorization

To extract files from Cloud Storage, establish valid authentication configurations:

1. Go to the GCP Console and navigate to IAM & Admin → Service Accounts.
2. Create a service account with the Storage Admin and Storage Object Viewer roles.
3. Whitelist the service account email within your target Cloud Storage bucket permissions panel.

2. Local Key Generation Fallback

If you encounter permission blockers or errors while downloading JSON keys from the console, authenticate locally via the Google Cloud CLI:

gcloud auth application-default login

This maps credentials locally to:

C:\Users\vigna\AppData\Roaming\gcloud\application_default_credentials.json

---

📊 Pipeline Orchestration

Data Ingestion & Preprocessing

• Run the ingestion module to extract the dataset from your GCP bucket and execute a structured train-test split.
• Add data_preprocessing parameters to config/config.yaml.
• Use preprocessing routines to balance target distribution flags using imbalanced-learn.

Experiment Tracking with MLflow

Configure model training hyperparameters inside config/model_params.yaml. To launch your experiment tracker and compare iterations, spin up the MLflow server:

mlflow ui

Dashboard URL: http://127.0.0.1:5000

---

🚀 CI/CD Automation via Jenkins & Cloud Run

The deployment pipeline relies on a custom Docker-in-Docker (DinD) Jenkins image to assemble runtime environments.

1. Build the Custom Jenkins Automation Image

cd custom_jenkins
docker build -t jenkins-dind .

2. Deploy the Jenkins Container

Launch your local automation server with exposed web management ports:

docker run -d --name jenkins-dind -p 8080:8080 -p 50000:50000 jenkins-dind:latest

3. Deployment Steps

1. Connect Jenkins to your GitHub repository webhook.
2. Configure your pipeline stage to login to Docker, assemble your Flask web app image, and push it directly to the Google Container Registry (GCR).
3. Extract the freshly built image from GCR and deploy it directly onto Google Cloud Run for public serverless hosting.

⚠️ Important: Ensure that the Artifact Registry API and Cloud Resource Manager API are enabled within your GCP Project console prior to executing the build pipeline.

---

📋 Prerequisites Summary

Tool	Purpose
Python 3.8+	Core runtime
Docker	Containerization & local Jenkins
Google Cloud SDK	GCP authentication & deployment
MLflow	Experiment tracking
Git	Large file / data versioning
Jenkins	CI/CD automation