Flight Altitude Prediction Method Based on Attention Mechanism and Convolutional Neural Network
MA Shan
GUO Wenxin
LI Fengming
YAN Dongfeng
SUN Xiaolin
Abstract：High-precision flight altitude prediction plays a decisive role in aviation safety operations and fuel efficiency en-hancement.This study develops a hybrid prediction model integrating temporal attention mechanisms with deep convolutional archi-tectures.First,normalization techniques are applied to reduce flight time-series signal data,followed by the design of a parallel net-work structure comprising multi-scale convolutional kernel feature extraction layers and attention weight allocation modules.The model significantly enhances its representation capability for nonlinear flight state variation patterns through gradient iterations of a Huber loss function optimized by a dynamically adjusted learning rate optimizer.Empirical analysis confirms the framework's excep-tional performance in altitude prediction tasks,achieving R2 coefficient exceeding 0.996 5.Ablation studies further substantiate the synergistic enhancement mechanism among model components.
Keywords：QAR dataattention mechanismconvolutional neural networkHuber losstiming prediction
Publication Date：2025-08-20
Online Publishing Date：2025-12-12(First online date of this platform, not the publication date of the document)
Pages：6( 38-43 )