A Multi-data Source Flight Departure Time Prediction Method Based on Sorting Learning

Abstract

The invention discloses a multi-data source flight departure time prediction method based on ranking learning, which comprises the following steps: using flight attributes and historical data of flight departure prediction time to perform prediction model training; optimizing the prediction model; The real-time data of the flight departure time of the data source is accepted. This method applies the ranking learning algorithm to the multi-data source decision-making of flight estimated departure time prediction, time-samples the historical data of flight prediction departure time based on multiple data sources, forms a flight document set by combining flight attributes, and based on the prediction error The flight departure time prediction is marked with relevance, and the ranking learning algorithm is called to obtain the predicted departure time with the highest score as the decision acceptance. The scheme of the present invention combines the historical prediction data of all data sources of the flight, reasonably utilizes the amount of prediction information, enriches the training data, and unifies the model to solve the comprehensive decision-making of the prediction and acceptance problem at any time in the entire life cycle of the flight.

Description

technical field [0001] The invention relates to a multi-data source flight departure time prediction method based on ranking learning, which belongs to the technical field of civil aviation information. Background technique [0002] Flight departure time prediction (ETD, estimated time of departure) determines the occupation of ground and airspace resources by civil aviation passenger aircraft, and is extremely important for the work efficiency of air traffic control, airports, airlines and other aviation units. For the same flight, the prediction of its departure time usually comes from multiple units. Because each data source has different resources, its expected coverage and error distribution are different, and often diverge. How to achieve high-quality comprehensive information decision-making is A problem worth solving. The acceptance scheme of a single data source can model the acceptance of a data source as a binary classification problem, and use common binary clas...

AI Technical Summary

Problems solved by technology

In the acquisition of training data, existing technologies are often limited to modeling separately within each data source

Images