XGBoost Hotel Booking Cancellation Prediction

This project provides a wrapper around an XGBoost model to predict hotel booking cancellations. It includes data preprocessing, one-hot encoding for categorical features, and prediction endpoints for both single and batch (file) inputs.

Features

• Predicts hotel booking cancellations using an XGBoost model.
• Handles both single prediction and batch predictions from CSV files.
• Preprocesses input data, including one-hot encoding of categorical features.
• Returns prediction probabilities and, for batch predictions, client information.

Project Structure

• src/wrapper.py: Main model wrapper for loading, preprocessing, and predicting.
• src/dtos.py: Data Transfer Objects (DTOs) for input and output schemas.
• type_adapter.py: (Contextual) Utilities for JSON schema and serialization.

Requirements

• Python 3.10+
• pandas
• xgboost
• scikit-learn

Install dependencies with:

pip install -r requirements.txt

Usage

Model Initialization

from src.wrapper import XGBoostModelWrapper

model = XGBoostModelWrapper('path/to/model.pkl', 'path/to/encoder.pkl')

Single Prediction

features = {
    "lead_time": 34,
    "hotel": "Resort Hotel",
    ...
}
result = model.predict_one(features)
print(result)

Batch Prediction from File

with open('input.csv', 'rb') as f:
    class FileObj:
        file = f
        filename = 'input.csv'
    result = model.predict_from_file(FileObj())
    print(result)

Input Format

• For single predictions, provide a dictionary matching the fields in FeaturesDto.
• For batch predictions, provide a CSV file with columns matching the features.

Output

• Single prediction: OnePredictionOutputDto with prediction and probabilities.
• Batch prediction: FilePredictionOutputDto with predictions, client info, and file metadata.

License

MIT License

Data example for one prediction

  "lead_time": 342,
  "arrival_date_year": 2015,
  "arrival_date_week_number": 27,
  "arrival_date_day_of_month": 1,
  "stays_in_weekend_nights": 0,
  "stays_in_week_nights": 0,
  "adults": 2,
  "children": 0,
  "babies": 0,
  "is_repeated_guest": 0,
  "previous_cancellations": 0,
  "previous_bookings_not_canceled": 0,
  "booking_changes": 3,
  "days_in_waiting_list": 0,
  "adr": 0,
  "required_car_parking_spaces": 0,
  "total_of_special_requests": 0,
  "hotel": "Resort Hotel",
  "arrival_date_month": "July",
  "meal": "BB",
  "country": "PRT",
  "market_segment": "Direct",
  "distribution_channel": "Direct",
  "reserved_room_type": "C",
  "assigned_room_type": "C",
  "deposit_type": "No Deposit",
  "customer_type": "Transient"
}

Docker commands:

For creating the container for the project use the following commands.

    docker build -t project-backend .
    docker run --name backend-project -p 8000:8000 -d project-backend