-
摘要: 为了解决交通检测器检测到的数据存在丢失的问题, 提出了一种基于粗集理论的丢失数据补齐方法。利用检测到的交通流数据构造信息系统, 通过计算扩充可辨识矩阵, 并对其进行多次完整化分析, 实施丢失数据的补齐, 并采用英国南安普敦市的实际检测数据对算法进行了验证。研究结果表明: 同一时间段, 当仅有一个属性数据丢失时, 粗集理论的补齐精度较高, 绝对相对误差较小, 基本保持在0~5%之间; 当不同属性的数据同时丢失时, 补齐精度较低, 绝对相对误差甚至高达20%;当所有属性数据全部丢失时, 补齐精度非常低, 可视为无法实现补齐。可见, 粗集理论是一种补齐少量丢失数据的有效方法。Abstract: In order to solve the problem of missing data detected from traffic detectors, the algorithm of filling missing data was proposed based on rough set theory, information system was constructed by the detected traffic flow data, distinct matrix was extended and analyzed repeatedly, the missing data of information system were filled, the algorithm was validated with the data of Southampton.Analysis result shows that, at the same period, when only one attribute datum is missed, the filling precision based on rough set theory is higher, the absolute relative error is lower, and basically keeps between 0 and 5%;when different attribute data are missed simultaneously, the filling precision is lower, and the absolute relative error is up to 20%;when all attribute data are missed, the filling is unable to realize, so the algorithm is very effective to fill a spot of missing data.
-
Key words:
- traffic engineering /
- traffic flow /
- missing data /
- rough set theory /
- filling method
-
表 1 信息系统S0
Table 1. Information system S0
U a1 a2 a3 U a1 a2 a3 U a1 a2 a3 x1 7 66 * x8 * 82 78 x15 * 78 75 x2 8 76 77 x9 8 * 71 x16 9 90 83 x3 5 47 80 x10 6 61 63 x17 * 85 80 x4 9 * 78 x11 9 68 * x18 11 80 75 x5 11 87 81 x12 9 71 70 x19 9 * 76 x6 9 73 75 x13 10 * 80 x20 10 82 * x7 10 80 77 x14 9 80 78 x21 10 85 80 表 2 信息系统S1
Table 2. Information system S1
U a1 a2 a3 U a1 a2 a3 U a1 a2 a3 x1 7 66 *68 x8 *10 82 78 x15 * 78 75 x2 8 76 77 x9 8 * 71 x16 9 90 83 x3 5 47 80 x10 6 61 63 x17 *10 85 80 x4 9 *80 78 x11 9 68 *76 x18 11 80 75 x5 11 87 81 x12 9 71 70 x19 9 *68 76 x6 9 73 75 x13 10 *85 80 x20 10 82 *78 x7 10 80 77 x14 9 80 78 x21 10 85 80 表 3 信息系统S2
Table 3. Information system S2
U a1 a2 a3 U a1 a2 a3 U a1 a2 a3 x1 7 66 *68 x8 *10 82 78 x15 *8 78 75 x2 8 76 77 x9 8 *76 71 x16 9 90 83 x3 5 47 80 x10 6 61 63 x17 *10 85 80 x4 9 *80 78 x11 9 68 *76 x18 11 80 75 x5 11 87 81 x12 9 71 70 x19 9 *68 76 x6 9 73 75 x13 10 *85 80 x20 10 82 *78 x7 10 80 77 x14 9 80 78 x21 10 85 80 表 4 数据对比
Table 4. Data comparison
时间 流量原始数据 算法补齐数据 绝对相对误差 时间 流量原始数据 算法补齐数据 绝对相对误差 时间 流量原始数据 算法补齐数据 绝对相对误差 07:25 61 61 0.000 11:35 66 66 0.000 15:40 80 79 0.012 08:30 54 66 0.222 11:55 80 80 0.000 16:10 122 120 0.016 09:10 73 82 0.123 12:50 80 80 0.000 16:20 88 88 0.000 09:35 70 73 0.042 13:15 71 76 0.070 17:45 108 105 0.028 09:40 113 108 0.044 13:35 85 85 0.000 17:50 111 111 0.000 10:20 89 83 0.067 14:05 81 68 0.160 19:40 68 69 0.015 10:40 99 87 0.121 14:35 87 87 0.000 11:05 77 74 0.039 14:55 83 83 0.000 -
[1] 王殿海. 交通流理论[M]. 北京: 人民交通出版社, 2002. [2] 王晓原, 隽志才, 贾洪飞, 等. 交通流突变分析的变点统计方法研究[J]. 中国公路学报, 2002, 15 (4): 69-74. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200204018.htmWANG Xiao-yuan, JUAN Zhi-cai, JI A Hong-fei, et al. Study of a statistical method of change-point to analyze trafficflow breakdown[J]. China Journal of Highway and Trans-port, 2002, 15 (4): 69-74. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL200204018.htm [3] 张敬磊, 王晓原. 交通事件检测算法研究进展[J]. 武汉理工大学学报: 交通科学与工程版, 2005, 29 (2): 215-218. https://www.cnki.com.cn/Article/CJFDTOTAL-JTKJ200502014.htmZHANG Jing-lei, WANG Xiao-yuan. Research progress oftraffic incident automatic detection algorithms[J]. Journal ofWuhan University of Technology: Transportation Scienceand Engineering, 2005, 29 (2): 215-218. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTKJ200502014.htm [4] 王晓原, 隽志才, 贾洪飞. 开发和评价ITS的微观交通流仿真模型[J]. 交通运输工程学报, 2002, 2 (1): 64-66. doi: 10.3321/j.issn:1671-1637.2002.01.014WANG Xiao-yuan, JUAN Zhi-cai, JIA Hong-fei. Micro-si mu-lation models of traffic flow of developing and evaluatingITS[J]. Journal of Traffic and Transportation Engineering, 2002, 2 (1): 64-66. (in Chinese) doi: 10.3321/j.issn:1671-1637.2002.01.014 [5] 王晓原, 隽志才, 贾洪飞, 等. 基于安全间距的车辆跟驰模型研究综述[J]. 长安大学学报: 自然科学版, 2004, 24 (6): 51-54. doi: 10.3321/j.issn:1671-8879.2004.06.013WANG Xiao-yuan, JUAN Zhi-cai, JI A Hong-fei, et al. Summarization of car-following models based on security dis-tance[J]. Journal of Chang an University: Natural ScienceEdition, 2004, 24 (6): 51-54. (in Chinese) doi: 10.3321/j.issn:1671-8879.2004.06.013 [6] 王晓原, 刘海红, 谭德荣. 交通流量变模式辨识的非参数概率变点模型[J]. 系统工程, 2006, 24 (8): 19-22. doi: 10.3969/j.issn.1001-4098.2006.08.004WANG Xiao-yuan, LI U Hai-hong, TAN De-rong. A non-parametric probability change-point model for traffic flowrecognition[J]. Systems Engineering, 2006, 24 (8): 19-22. (in Chinese) doi: 10.3969/j.issn.1001-4098.2006.08.004 [7] 王晓原, 杨新月. 驾驶行为非参数微观仿真模型[J]. 交通运输工程学报, 2007, 7 (1): 76-80. http://transport.chd.edu.cn/article/id/200701017WANG Xiao-yuan, YANG Xin-yue. Nonparametric micro-scopic si mulation model of driving behavior[J]. Journal ofTraffic and Transportation Engineering, 2007, 7 (1): 76-80. (in Chinese) http://transport.chd.edu.cn/article/id/200701017 [8] 耿彦斌, 于雷, 赵慧. ITS数据质量控制技术及应用研究[J]. 中国安全科学学报, 2005, 15 (1): 82-87. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK200501021.htmGENG Yan-bin, YU Lei, ZHAO Hui. ITS data quality con-trol techniques and applications[J]. China Safety ScienceJournal, 2005, 15 (1): 82-87. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK200501021.htm [9] 陈德旺, 郑长青, 章长彪. 快速路交通流异常数据判断算法研究及实证[J]. 中国安全科学学报, 2006, 16 (7): 122-127. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK200607024.htmCHEN De-wang, ZHENG Chang-qing, ZHANG Chang-biao. An algorithmfor judging abnormal data of expressway trafficflow and its validation[J]. China Safety Science Journal, 2006, 16 (7): 122-127. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK200607024.htm [10] 姜桂艳, 冮龙晖, 张晓东, 等. 动态交通数据故障识别与修复方法[J]. 交通运输工程学报, 2004, 4 (1): 121-125. http://transport.chd.edu.cn/article/id/200401030JI ANG Gui-yan, GANG Long-hui, ZHANG Xiao-dong, et al. Malfunction identifying and modifying of dynamic trafficdata[J]. Journal of Traffic and Transportation Engineering, 2004, 4 (1): 121-125. (in Chinese) http://transport.chd.edu.cn/article/id/200401030 [11] 王国胤. Rough集理论与知识获取[M]. 西安: 西安交通大学出版社, 2001. [12] HONG T P, TSENG L H, WANGS L. Learning rules fromincomplete training examples by tough sets[J]. Expert Sys-tems with Applications, 2002 (22): 285-293.