基于轻量化网络MobileNetV2的航班延误预测模型

Flight Delay Prediction Model Based on Lightweight Network MobileNetV2

  • 摘要: 针对民航业中航班延误状况的日益凸显,传统算法存在准确率低、计算量以及参数量大的问题,且面对旅客主要使用移动设备查询的需求,传统算法难以直接部署在移动端,本文提出一种基于轻量化网络MobileNetV2的航班延误预测模型。模型首先对数据集做数据融合、编码等预处理;然后将其输入到网络中进行特征提取;最后利用Softmax分类器输出航班延误等级。应用于国内数据集,准确率最高为99.07%,模型参数量为1.31Million、计算量为40.58Million。本文模型在保障准确率的同时,尽可能降低模型的参数量和计算量,其性能优于传统网络,有助于在移动端实现航班延误预测。

     

    Abstract: ‍ ‍For the increasing prominence of flight delay in civil aviation industry, the traditional algorithms have problems with low accuracy, calculation quantity and large number of parameters. Facing the demand of passengers' main query using mobile devices, traditional algorithms are difficult to be directly deployed on the mobile terminal. A flight delay prediction model based on lightweight network MobileNetV2 is proposed in this paper. Firstly, Datasets are to be fused and encoded in this model. Then, the preprocessed data is input into the network for feature extraction. Finally, it uses the Softmax classifier to output the flight delay level. Applied to domestic datasets, the highest accuracy is 99.07% and the number of parameters of the model is 1.31Million, computation of 40.58Million. While ensuring the accuracy, the present model reduces the parameters and computation of the model, which outperforms the traditional network, and helps to realize the flight delay prediction on the mobile end.

     

/

返回文章
返回