基于BiLG-A-CNN的冷链配送公司评价模型

BiLG-A-CNN Based Evaluation Model For Cold Chain Distribution Companies

随着互联网技术的发展,网上购物成为一种大众消费项目,促生了冷链物流产业的大力发展,然而,冷链配送公司良莠不齐。为了便于贵州生鲜电商人员选择最优的冷链配送公司,选取了贵州省500家冷链配送公司作为研究对象,并选取占地面积、月行驶里程、机构设置等18个指标作为评价指标,构建了基于贵州省的冷链配送公司评价模型。首先,模型使用平行的双向长短期记忆(BiLSTM)网络和双向门控循环单元(BiGRU)网络同时记忆短和长的时序特征;然后,模型通过注意力机制(Attention)为不同的特征分配不同的关注度;最后,模型通过卷积神经网络(CNN)提取局部特征和降维。通过多组对比实验可以发现,所建模型的准确(Accuracy)、精确(Precision)、召回(Recall)以及F1分数(F1-score)值分别达到97.75%、98.01%、97.54%以及98.12%,这一结果远高于对比模型。

With the development of Internet technology,online shopping has become a kind of mass consumption project,which has promoted the vigorous development of cold chain logistics industry.However,cold chain distribution companies are mixed.In order to facilitate Guizhou fresh e-commerce personnel to choose the optimal cold chain distribution company,500 cold chain distribution companies in Guizhou province are selected as the research object,and 18 indicators such as floor space,monthly driving mileage and institutional settings are selected as evaluation indexes,and the evaluation model of cold chain distribution companies based on Guizhou province is constructed.First,the model uses a parallel Bidirectional Long and Short-Term Memory(BiLSTM) network and a Bidirectional Gated Recurrent Unit(BiGRU) network to memorize both short and long temporal features;then,the model assigns different attention to different features through an Attention mechanism;finally,the model extracts local features and dimensionality reduction through a Convolutional Neural Network(CNN).Comparison experiment results show that the Accuracy,Precision,Recall,and F1-score values of the proposed model reach 97.75%,98.01%,97.54%,and 98.12%,respectively,which is much higher than that of the comparison model.