| 引用本文: | 付旻,王炜,王昊,项昀.多分类支持向量机在公交换乘识别中的应用[J].哈尔滨工业大学学报,2018,50(3):26.DOI:10.11918/j.issn.0367-6234.201610057 |
| FU Min,WANG Wei,WANG Hao,XIANG Yun.Application of multi-class support vector machine in public transit transfer recognition[J].Journal of Harbin Institute of Technology,2018,50(3):26.DOI:10.11918/j.issn.0367-6234.201610057 |
|
| 本文已被：浏览 1691次 下载 921次 |
码上扫一扫！ |
|
|
| 多分类支持向量机在公交换乘识别中的应用 |
|
付旻1,2,王炜1,2,王昊1,2,项昀1,2,3
|
|
(1. 城市智能交通江苏省重点实验室(东南大学),南京 210096; 2. 现代城市交通技术江苏高校协同创新中心(东南大学),南京 210096;3. 南昌航空大学 土木建筑学院,南昌 330063)
|
|
| 摘要: |
| 为获取居民公交出行的换乘信息，设计了一套基于多分类支持向量机(multi-class support vector machine)的公交换乘识别方法.通过融合GPS数据和公交IC卡数据获取训练样本，利用多分类支持向量机进行样本训练，选取最佳训练样本量，并采用网格搜索法结合粒子优化算法对模型参数进行标定，以获取最优SVM分类模型.测试结果显示模型分类精度可达90%.以佛山市公交车GPS数据和IC卡数据对算法进行验证，并获取公交换乘量、公交换乘比例等基本换乘数据.结果表明：算法可在少样本条件下完成公交换乘识别，且分类识别精度高，尤其适用于公交线网复杂的大城市公交换乘识别，有助于在公交前期规划时进行线路布设和枢纽选址.
|
| 关键词: 公交换乘识别 公交GPS和IC卡数据 同站换乘 异站换乘 多分类支持向量机 |
| DOI：10.11918/j.issn.0367-6234.201610057 |
| 分类号:U491.112 |
| 文献标识码:A |
| 基金项目:国家自然科学基金重点资助(51338003) |
|
| Application of multi-class support vector machine in public transit transfer recognition |
|
FU Min1,2,WANG Wei1,2,WANG Hao1,2,XIANG Yun1,2,3
|
|
(1. Jiangsu Key Laboratory of Urban ITS (Southeast University), Nanjing 210096, China; 2. Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic Technologies (Southeast University), Nanjing 210096, China; 3. School of Civil Engineering and Architecture, Nanchang Hangkong University, Nanchang 330063, China)
|
| Abstract: |
| To obtain the transfer information of passengers' public transit behavior, a public transit transfer recognition method is designed based on multi-class support vector machine (multi-class SVM). GPS data and intelligent card data are fused to get sufficient samples, then the Multi-class support vector machine model is used to train the samples. The best sample size could be acquired by accuracy control, and the Grid-Search method combined with Particle Swarm Optimization method is employed to determine the parameter for gaining the optimal SVM model. Finally, a case study with GPS data and intelligent card data in Foshan city is conducted to verify the algorithm, this method can acquire transfer characteristics including transfer flow and transfer proportion etc. Results show that the proposed method could complete public transit transfer recognition with high classification accuracy even if the size of training sample is rather small. Especially, it is useful for transfer recognition in large cities with complex public transit networks, which provides a basis for public transit lines planning and pub selection.
|
| Key words: public transit transfer recognition GPS and intelligent card data transfer inside one station transfer between different stations multi-class support vector machine |
