Flight Delay Prediction Model Based on Lightweight Network ECA-MobileNetV3

In exploring the flight delay problem, traditional deep learning algorithms suffer from low accuracy and extreme computational complexity; therefore, prediction algorithm is difficult to directly deploy mobile terminal. this paper, a model based on lightweight network ECA-MobileNetV3 proposed. The first preprocesses data with real information weather information. Then, in order increase of without increasing complexity too much, feature extraction performed using addition Efficient Channel Attention mechanism. Finally, classification level output via Softmax classifier. experiments single airport cluster datasets, optimal 98.97% 96.81%, number parameters 0.33 million 0.55 million, volume 32.80 60.44 respectively, which are better than performance MobileNetV3 under same conditions. improved can achieve balance between complexity, more conducive mobility.