| 引用本文: | 方琼,钱大琳,陈心如,李思贤.危险货物道路运输个性化路径推荐方法[J].哈尔滨工业大学学报,2024,56(7):55.DOI:10.11918/202211034 |
| FANG Qiong,QIAN Dalin,CHEN Xinru,LI Sixian.Personalized route recommendation method for road transport of hazardous materials[J].Journal of Harbin Institute of Technology,2024,56(7):55.DOI:10.11918/202211034 |
|
| |
|
|
| 本文已被：浏览 149次 下载 201次 |
码上扫一扫！ |
|
|
| 危险货物道路运输个性化路径推荐方法 |
|
方琼1,钱大琳1,陈心如2,李思贤1
|
|
(1.综合交通运输大数据应用技术交通运输行业重点实验室(北京交通大学),北京 100044; 2.武汉地铁运营有限公司,武汉 430035)
|
|
| 摘要: |
| 为加强危险货物道路运输风险源头管控,以危货运输车辆行驶轨迹数据为分析对象,研究安全、经济且符合企业自身偏好的道路运输路径优化选择问题,提出了基于偏好、上下文感知的危险货物道路运输个性化路径推荐方法。首先对危货运输车辆历史轨迹数据进行处理,通过提取路径安全和经济性特征学习危货运输企业的路径偏好,在此基础上,综合考虑偏好向量间的距离和方向相似性,提出了改进的K-means偏好聚类算法(improved K-means clustering algorithm based on distance and direction similarity measurement,DDM-K-means),获取了路径偏好类别；其次,依据运输任务执行的时间、天气、运距三方面信息,建立了路径上下文向量,并运用Rock聚类算法划分路径的上下文类别,与路径偏好类别共同构成路径信息；最终,基于神经协同过滤提出了危险货物道路运输路径选择优化算法(optimal route selection algorithm based on neural collaborative filtering,NCF-ORS),得到了危货运输企业对各路径类别的偏好排序,从而为企业推荐最优路径。与基线算法比较分析,结果表明危险货物道路运输个性化路径推荐方法,平均绝对百分比误差最低。研究结果有助于挖掘车辆轨迹数据中更多的潜在信息,提升个性化路径推荐能力,可为危货运输企业的选线问题提供决策支持。 |
| 关键词: 危险品运输 路径推荐 神经协同过滤 偏好聚类算法 上下文感知 |
| DOI：10.11918/202211034 |
| 分类号:U492.336.3 |
| 文献标识码:A |
| 基金项目:国家自然科学基金(0,9) |
|
| Personalized route recommendation method for road transport of hazardous materials |
|
FANG Qiong1,QIAN Dalin1,CHEN Xinru2,LI Sixian1
|
|
(1.Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport (Beijing Jiaotong University), Beijing 100044, China; 2.Wuhan Metro Operation Co., Ltd., Wuhan 430035, China)
|
| Abstract: |
| To strengthen the management and control of the source of risk for road transport of hazardous materials, this paper takes the trajectory data of hazardous materials transport vehicles as the analysis object, and studies the problem of optimal selection of road transport routes which is safe, economical and in line with the preference of enterprises, a context-aware, preference-based personalized route recommendation method for road transport of hazardous materials is proposed. Firstly, the historical trajectory data of hazardous materials transport vehicles is processed, and the route preferences of enterprises are learned by extracting route safe and economical features. On this basis, considering the distance and direction similarity between preference vectors, an improved K-means clustering algorithm (DDM-K-means) is proposed to obtain the categories of route preference. Secondly, according to the time, weather, and distance of the transportation tasks, the route context vectors are established. Rock clustering algorithm is used to classify the categories of route context, combined with the categories of route preference to form the categories of route. Finally, based on neural collaborative filtering, an optimal route selection algorithm (NCF-ORS) is proposed, and the preference ranking of hazardous materials road transport enterprises for route categories is obtained to recommend the optimal route for enterprises. Comparing our method with the baseline algorithms, the results showed that the personalized route recommendation method proposed in this paper had a lower mean absolute percentage error. Therefore, the research in this paper is helpful to mine more potential information from vehicle′s trajectory data, with stronger personalized route recommendation capabilities, and can provide decision support for route selection of hazardous materials road transport enterprises. |
| Key words: transportation of hazardous materials route recommendation neural collaborative filtering preference clustering algorithm context-aware |
|
|
|
|
