🏨 Hotel Reservation Prediction

This project aims to predict whether a hotel booking will be canceled based on various customer and reservation-related features. By leveraging machine learning techniques and modern deployment practices, the solution enables proactive management of hotel reservations to improve operational efficiency.

---

🔍 Problem Statement

Hotel booking cancellations lead to significant revenue losses and poor resource planning. This project helps hotels predict the likelihood of a booking being canceled based on customer behavior, booking patterns, and preferences.

---

📊 Features Used

Categorical Columns

• type_of_meal_plan
• required_car_parking_space
• room_type_reserved

Numerical Columns

• no_of_adults
• no_of_children
• no_of_weekend_nights
• no_of_week_nights
• lead_time
• arrival_date
• avg_price_per_room
• no_of_special_requests

---

🧠 ML Workflow

1. Data Infrastructure Setup

• Data stored and maintained in a Google Cloud Storage (GCP Bucket).

2. Development Environment

• Codebase built using VS Code with Python.
• Versioning via GitHub.

3. Data Ingestion

• Data pulled directly from GCP Bucket into local development environment.

4. Exploratory Data Analysis (EDA) & Preprocessing

• Univariate & Bivariate Analysis to explore data trends.
• Label Encoding for categorical features.
• Variance Inflation Factor (VIF) to detect multicollinearity.
• Handling skewness, imbalanced datasets.
• Feature selection using RandomForest importance scores.

5. Model Selection

Models compared based on key metrics:

• ✅ Accuracy
• 📌 Precision
• 🔁 Recall
• 🎯 F1 Score

Models Tested:

• Random Forest (RF)
• Gradient Boosting
• AdaBoost
• Logistic Regression
• Support Vector Classifier (SVC)
• K-Nearest Neighbors (KNN)
• Gaussian Naive Bayes (NB)
• LightGBM (LGBM)
• XGBoost (XGB)

✅ Best Model: Random Forest Classifier

6. Hyperparameter Tuning

• Using RandomizedSearchCV to optimize:
  • bootstrap
  • max_depth
  • min_samples_leaf
  • min_samples_split
  • n_estimators

7. Model Training & Saving

• Final model trained and serialized using joblib.

8. Experiment Tracking

• Integrated MLFlow for tracking experiments, metrics, and models.

9. Pipeline Development

• Modular training pipeline created for scalability and reusability.

10. Version Control

• Code Versioning: Git + GitHub
• Data Versioning: GitHub + GCP bucket logging

---

🌐 User Interface

Flask Application

• Built using Flask to interact with the trained model.

• Accepts user input and predicts if the booking will be canceled.

  

---

🚀 CI/CD Deployment

Containerization & Deployment Workflow:

1. Create virtual environment using Jenkins pipeline.

   

2. Dockerize the project.

3. Build Docker image and push to Google Cloud Container Registry (GCR).

   

4. Deploy image to Google Cloud Run for seamless scalability.

   

---

🛠️ Testing Application:

🛠️ Tech Stack

• Python 🐍
• Scikit-learn, Pandas, NumPy, Matplotlib, Seaborn
• MLFlow for experiment tracking
• Flask for UI/API
• Docker, Jenkins for CI/CD
• Google Cloud Platform (GCS, GCR, Cloud Run)

---

📦 Dataset

The dataset used for this project can be accessed here on Kaggle.

---

🙏 Acknowledgements

This project is part of a course taught by Sudhanshu Gusain on udemy.com.
Thanks to the instructor for a practical, hands-on approach to solving real-world ML problems!

---