Flight Price Prediction
✈️ Flight Price Prediction
Utilizing machine learning to predict airfare prices with precision, this project provides insights into airfare trends based on factors like airline, travel class, departure time, and more.
📂 GitHub Repository
📌 Overview
Flight prices fluctuate due to multiple factors, making prediction a complex task. This project leverages advanced machine learning models to forecast ticket prices, helping users make informed travel decisions.
📊 Dataset
The dataset includes the following key attributes:
- Airline: Name of the airline
- Source City: City of departure
- Destination City: City of arrival
- Departure Time: Categorized into Morning, Afternoon, Evening, etc.
- Arrival Time: Categorized into Morning, Afternoon, Evening, etc.
- Stops: Number of stops during the journey
- Class: Travel class (Economy or Business)
- Price: Ticket price (Target variable)
🔎 Process
To enhance model performance, the following steps were taken:
🛠️ Data Preprocessing
- Data Cleaning:
- Merged Morning & Early Morning into a single category.
- Merged Night & Late Night into a single category.
- Encoding: Applied One-Hot Encoding to categorical columns like airline, source city, and destination city.
- Scaling: Used MinMaxScaler to normalize the target column (Price).
- Feature Selection: Retained important features such as stops, class, and departure time.
🚀 Model Training
Multiple machine learning models were explored:
- Gradient Boosting
- Random Forest Regressor
- XGBoost Regressor
Hyperparameter tuning was conducted using RandomizedSearchCV to optimize performance.
🏆 Best Model: XGBoost
XGBoost was chosen based on its superior performance in terms of evaluation metrics:
✅ Model Performance
| Metric | Score |
|---|---|
| Mean Squared Error (MSE) | 0.0464 |
| Root Mean Squared Error (RMSE) | 0.2156 |
| R² Score | 0.95 |
📌 Feature Importance Analysis
The most influential features impacting flight prices were:
✔ Stops
✔ Class
✔ Departure Time
🔍 Future Improvements
- Feature Engineering: Explore additional features such as holiday seasons, demand surges, and ticket booking windows to improve accuracy.
- Deep Learning Models: Implement LSTMs or Transformer-based models for better sequential trend analysis.
- Real-Time Price Tracking: Integrate real-time API data to enhance model predictions based on current trends.
- Hyperparameter Tuning: Further optimize hyperparameters using Bayesian Optimization for better performance.
- Explainability: Use SHAP values to better interpret model decisions and improve trust in predictions.
🚀 Continuous improvements will ensure more precise and reliable flight price predictions!
🚀 This project showcases how machine learning can help predict airfare trends, aiding travelers in making cost-effective booking decisions!