Selection of Best Machine Learning Model to Predict Delay in Passenger Airlines

Aim:

         The aim of this project is to develop a flight delay prediction system that uses a decision tree algorithm to predict delays based on historical data while providing real-time flight tracking and delay updates using Neo4j and live APIs. The system will be deployed on a web interface to provide users with accurate and timely flight delay predictions.

Abstract:

        Flight delays are a persistent issue in the aviation industry, largely due to growing air traffic and operational challenges. This project proposes a predictive model using machine learning techniques to estimate flight delays based on historical data and real-time flight information. Our model leverages a decision tree algorithm to deliver accurate delay predictions, with live flight data integration visualized through Neo4j. The system is designed for deployment via a web interface, offering an intuitive platform for users to track flights and anticipate delays between any source and destination.

Existing System:

        The current systems for flight delay prediction often rely on machine learning models, such as random forests. These systems focus on analysing historical data, considering factors like weather, traffic, and operational delays. Although these models provide reasonable accuracy, they tend to lack real-time data integration and live tracking capabilities, which limits their effectiveness in predicting delays as conditions change.

Disadvantages:

Inconsistent Real-time Updates: Existing models often do not incorporate real-time flight data, limiting their responsiveness to current conditions.

Complexity of Models: Techniques like random forests can be computationally intensive and difficult to interpret, making it challenging for non-experts to understand the underlying factors of delay predictions.

Scalability Issues: Many current models struggle to handle the volume of real-time data generated from numerous flights, leading to potential delays in prediction and updates.

Proposed System:

        In the proposed system, we use a decision tree algorithm, which offers a simpler yet effective approach for predicting flight delays with improved accuracy. This model outperforms the existing systems by providing real-time updates through live flight tracking integrated with Neo4j. Our web-based platform allows users to input flight details (source, destination) and receive real-time delay predictions, using historical and live flight data. The decision tree model provides better interpretability compared to more complex models and can achieve accuracy on par with existing systems while being faster and easier to deploy.