Scalable flight cancellation prediction with ensemble distributed KNN and feature selection.

Area of Science:

• Aviation Data Science
• Machine Learning in Transportation
• Predictive Analytics

Background:

• Accurate flight cancellation prediction is crucial for airline financial health and passenger satisfaction.
• Traditional methods struggle with the volume and complexity of aviation big data.
• Existing prediction models lack scalability and optimal feature selection capabilities.

Purpose of the Study:

• To develop an innovative distributed ensemble learning approach for large-scale flight cancellation prediction.
• To enhance prediction accuracy and computational efficiency using big data environments.
• To address the limitations of traditional prediction methods in the aviation industry.

Main Methods:

• Utilized distributed ensemble learning with the MapReduce framework for scalability.
• Implemented distributed K-Nearest Neighbor (DKNN) models for efficient data processing.
• Employed the Artificial Bee Colony (ABC) algorithm for optimal feature selection from extensive datasets.

Main Results:

• Achieved a minimum computational advantage exceeding 25% compared to non-distributed KNN models.
• The ensemble strategy improved prediction accuracy by 3.42%, reaching an average of 95.79%.
• Demonstrated a 2.2% improvement in accuracy over previous flight cancellation prediction methods.

Conclusions:

• The proposed distributed ensemble methodology effectively predicts flight cancellations in big data environments.
• The system offers significant computational advantages and enhanced prediction accuracy.
• This approach provides a scalable and accurate solution for the aviation industry's prediction challenges.