This project develops a flight price prediction model for the Economy and Business segments, applying a complete supervised Machine Learning workflow, from data exploration to model evaluation.

The analysis aims to identify the factors that most strongly influence flight prices and to build models capable of anticipating fares based on variables such as airline, origin, destination, flight duration, and number of stops.

The project integrates an analytical, technical, and business-oriented perspective, focused on improving the understanding of the airline market and supporting dynamic pricing strategies and commercial optimization.

• Data cleaning and outlier detection.
• Univariate and bivariate analysis of key variables.
• Study of correlations and price distributions.

• Generation of time-based variables (month, day, weekday).
• Categorical encoding using One-Hot Encoding.
• Selective scaling of numerical variables with MinMaxScaler.

Comparison of three representative regression approaches:

• Ridge Regression: linear baseline model with L2 regularization.
• Random Forest Regressor: tree-based ensemble capturing non-linear relationships.
• Gradient Boosting Regressor: sequential model optimized on residual errors.

The evaluation metrics used were MAE, RMSE, and R².

• Random Forest achieved the best balance between accuracy and generalization.
• Flight duration, airline, and origin–destination routes showed the strongest influence on final ticket prices.

Cristián Andrés Galleguillos Vega
Biologist · MSc in Natural Resources Engineering · MSc in Data Science & Big Data
Location: Chile
LinkedIn: Cristián Galleguillos Vega

This project demonstrates the practical application of Machine Learning for price prediction in a real-world business context.
The proposed analytical workflow is reproducible, interpretable, and scalable, making it a solid foundation for future projects in price optimization, demand forecasting, or consumer behavior analysis.