SVM Flight Delay Prediction Model Using Prior Knowledge
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing flight delay prediction methods focus primarily on short-time delay rates and average delay times, failing to accurately predict flight operation situations over extended time periods, which are influenced by multiple random factors such as weather and airline operations.
Innovation Solution
A method using flight historical data to determine prior knowledge and construct a Support-Vector-Machine (SVM) prediction model, which is verified and updated using Massive Delay Response System (MDRS) information to improve accuracy and reliability of flight delay predictions across various time periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional statistical learning methods are used to predict flight delay, then the prediction of short-time delay rate and average delay time can be achieved, but the prediction accuracy for extended time periods deteriorates due to the high randomness of multiple influencing factors
Solution Approach 1:
The patent applies preliminary action by determining prior knowledge for each time period before constructing the SVM prediction model. This prior knowledge is derived from historical flight data and is used to pre-configure the prediction model, enabling it to adapt to different time periods' characteristics before actual prediction, thereby improving both accuracy and reliability across extended time periods.
Solution Approach 2:
The patent implements dynamics by constructing a time-period-specific SVM prediction model where the prior knowledge parameters are dynamically determined for each time period based on historical data patterns. This allows the model to adapt its prediction behavior to the specific characteristics of different time periods, rather than using a static model for all periods.
2Measurement precision
If prior knowledge is incorporated into the SVM model for each time period, then the prediction accuracy improves, but the model complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the prediction task into multiple time periods, with each period having its own prior knowledge determination and SVM model construction. This segmentation allows the complex problem of extended time period prediction to be broken down into manageable segments, where prior knowledge is determined and models are constructed for each segment individually.
Solution Approach 2:
The patent implements parameter changes by determining different prior knowledge parameters for each time period based on historical flight data. These parameter changes allow the SVM model to adapt to different time period characteristics without requiring a completely different model structure, thus improving accuracy while controlling complexity through parameter adjustment rather than structural complexity.
Data Source
AI summary
The disclosure provides a method and an apparatus for predicting flight delay, a device and a storage medium. The method includes: acquiring flight historical data, where the flight historical data includes take-off amount and delay amount of flights during each of a plurality of time periods; determining prior knowledge of each of the plurality of time periods according to the take-off amount and the delay amount of the flights during each of the plurality of time periods; constructing a SVM prediction model according to the prior knowledge and a standard SVM model; and predicting a flight delay situation according to the SVM prediction model. The prediction of the flight delay situation for each of the plurality of time periods is realized.