摘要
随着网约车的蓬勃发展,网约车为我们的出行带来了极大的便利。而随着市场规模的逐步扩大,网约车安全日益成为网约车行业关注的焦点。研究尝试通过机器学习辅以文本挖掘手段来对司乘冲突的严重程度进行定级预测,数据来源于法律文书网和资讯网站。研究发现在数据特征选择方面,机器学习能选择出更好的特征变量,而当采用特征标准化方法处理样本并且采用逻辑回归作为模型算法时,网约车司乘冲突事态严重程度等级预测准确率最高。研究发现,冲突事态严重程度定级对于网约车安全事件管理有着重要理论意义,另一方面,网约车平台通过机器学习手段建立主动发现冲突风险的主动防御型风险管理模式,以此降低司乘冲突风险来提升网约车司乘双方的安全,具备一定的实践意义。
With the vigorous development of online car hailing,online car hailing has brought great convenience for our travel.However,with the gradual expansion of the market scale,the safety of online car hailing has increasingly become the focus of online car hailing industry.This study attempts to use machine learning and text mining to predict the severity of conflict between drivers and passengers.It is found that machine learning can select better feature variables in data feature selection,and when using feature standardization method to process samples and using logistic regression as model algorithm,the prediction accuracy of the severity level of the conflict between drivers and passengers is the highest.Through the study of this paper,the classification of conflict severity has important theoretical significance for the safety incident management of online car hailing.On the other hand,the online car hailing platform can establish an active defense risk management mode to actively discover the conflict risk through machine learning,so as to reduce the conflict risk of drivers and passengers and improve the safety of both drivers and passengers,which has certain practical significance.
作者
余琴
侯立文
YU Qin;HOU Liwen(Shanghai Jiao Tong University,Shanghai 200030,China)
出处
《上海管理科学》
2023年第5期44-50,共7页
Shanghai Management Science
关键词
网约车安全
司乘冲突
严重性分级
机器学习
特征工程
car-hailing safety
conflict between drivers and passengers
severity grade
machine learning
feature engineering