基于深度学习的地铁站点环境满意度评估算法

Deep learning-based algorithm for environmental satisfaction assessment at metro stations

为全面评估北京地铁乘客对站内环境的满意度,对地铁站内环境图像进行经过隐私保护处理的数据采集,通过问卷调查和随机森林算法,确定影响地铁站内环境满意度的关键指标及其权重,利用被试人员的评估数据,建立包含满意度评分的环境满意度评估数据集。通过学习环境图像与满意度评分的关系,构建深度卷积神经网络模型。为增强模型泛化性,采用迁移学习方法。在北京地铁14号线测试集上,模型评估准确率达到96%,在未经预训练的地铁7号线上迁移学习模型达到94%的准确率。实验结果表明,该模型可准确反映乘客对地铁环境在可达性、安全性、舒适性和愉悦性方面的体验,为地铁服务的优化提供有力参考。

To comprehensively assess the passenger satisfaction of the environment in Beijing metro stations,firstly,the data collection of the environment images in metro stations with privacy-protected processing is carried out,and through questionnaire surveys and random forest algorithms,the key indexes and their weights affecting the satisfaction of the environment in metro stations are determined,and the assessment data of the subjects are utilized to establish the environmental satisfaction assessment dataset which contains the satisfaction scores.Next,a deep convolutional neural network model is constructed by learning the relationship between environmental images and satisfaction scores.To enhance the model generalization,migration learning method is used.The model assessment accuracy reaches 96%on the test set of metro line 14,and the migration learning model achieves 94%accuracy on the metro line 7 which is not pretrained.The experimental results show that the model can accurately reflect passengers'experience of the metro environment in terms of accessibility,safety,comfort and pleasantness,providing a powerful reference for the optimization of metro services.