Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2021, Vol. 55 Issue (6): 1072-1082    DOI: 10.3785/j.issn.1008-973X.2021.06.007
    
Model based on relevance vector machine and fuzzy comprehensive evaluation for road condition prediction
Hao LIN1,2(),Lei-xiao LI1,2,*(),Hui WANG1,2,Zhi-qiang MA1,2,Jian-xiong WAN1,2
1. College of Data Science and Application, Inner Mongolia University of Technology, Hohhot 010080, China
2. Inner Mongolia Autonomous Region Engineering and Technology Research Center of Big Data Based Software Service, Hohhot 010080, China
Download: HTML     PDF(1303KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

A model based on relevance vector machine and fuzzy comprehensive evaluation was proposed in order to solve the problems of traffic congestion, low road service level, and low efficiency of citizens’ travel. Genetic algorithm and particle swarm optimization algorithm were used as the parameter optimization algorithm in order to optimize relevance vector machine with combined kernel functions. Then the parameter optimization algorithm was parallelized by Spark to improve the training efficiency of model. Genetic algorithm and particle swarm optimization based on Spark parallelization were proposed, and combined kernel relevance vector machine (SPGAPSO-CKRVM) was optimized. Traffic flow and traffic speed were predicted by SPGAPSO-CKRVM, and the prediction results were used to calculate three traffic condition evaluation parameters: average speed, road occupancy and traffic density. These three parameters were input into fuzzy comprehensive evaluation model. Weight coefficients of these evaluation parameters in peak hours and normal hours were determined by entropy method. Road condition was divided into six levels. The proposed model was verified with the real data of Whitemud Drive in Canada. The experimental results show that the proposed model has better prediction accuracy and scalability than traditional methods. The accuracy of road condition prediction can reach 90.28%.

Key words： road condition prediction      relevance vector machine (RVM)      combined kernel function      fuzzy comprehensive evaluation      Spark     
Received: 02 July 2020      Published: 30 July 2021
CLC:  TP 181  
Fund:  内蒙古自治区科技重大专项资助项目（2019ZD015）；内蒙古自治区关键技术攻关计划资助项目（2019GG273）
Corresponding Authors: Lei-xiao LI     E-mail: suzukaze_aoba@foxmail.com;llxhappy@126.com
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Hao LIN
Lei-xiao LI
Hui WANG
Zhi-qiang MA
Jian-xiong WAN
Cite this article:

Hao LIN,Lei-xiao LI,Hui WANG,Zhi-qiang MA,Jian-xiong WAN. Model based on relevance vector machine and fuzzy comprehensive evaluation for road condition prediction. Journal of ZheJiang University (Engineering Science), 2021, 55(6): 1072-1082.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2021.06.007     OR     https://www.zjujournals.com/eng/Y2021/V55/I6/1072


基于相关向量机和模糊综合评价的路况预测模型

为了缓解交通拥堵、道路服务水平低、市民出行效率低等问题，提出基于相关向量机和模糊综合评价的路况预测模型. 利用遗传算法和粒子群算法作为参数寻优算法，优化组合核相关向量机. 基于Spark并行化参数寻优算法，提高模型的训练效率. 提出基于Spark并行化的遗传算法和粒子群算法，优化组合核相关向量机（SPGAPSO-CKRVM）. 使用SPGAPSO-CKRVM对车流量和车速进行预测，利用预测结果计算3个交通路况评价参数：平均车速、路段饱和度和交通流密度. 将3个参数输入到模糊综合评价模型中，通过熵值法确定高峰时段和平常时段的各指标权重系数，将路况划分为6个等级. 使用加拿大Whitemud Drive公路的真实数据进行验证，证明了该模型与传统方法相比具有更高的预测精度和扩展性，路况预测准确率达到90.28%.


关键词： 路况预测,  相关向量机（RVM）,  组合核函数,  模糊综合评价,  Spark 
Fig.1 Flow figure of proposed road prediction model
Fig.2 Membership function used in this paper
Fig.3 Structure of parameter optimization algorithm
Fig.4 Time consumed by each part in parameter optimization algorithm
Fig.5 Algorithm flow chart of SPGAPSO-CKRVM
数据集编号 道路 方向 监测点编号
数据集1 Whitemud Drive 向东 1027
数据集2 Whitemud Drive 向西 1036
数据集3 Whitemud Drive匝道 向西 1042
Tab.1 Information of experiment data in Whitemud Drive Road
参数 参数值
种群大小 $m$ 10
最大迭代次数 $T$ 20
最小适应度 ${\rm{fitnes}}{{\rm{s}}_{{\rm{min}}}}$ 0.000 1
交叉概率 0.6
变异概率 0.2
粒子群学习因子 1.5
粒子速度 [?0.2，0.2]
粒子位置 [?8，8]
Tab.2 Parameter setting of SPGAPSO-CKRVM
编号 核函数 准确率
数据集1 数据集2 数据集3
1 $\exp\; \left( { - \dfrac{ {\left\| {{{a}}_1 - {{{a}}_2} } \right\|} }{ { {\rm{2} }{\sigma ^{\rm{2} } } } } } \right)$ 0.8084 0.7964 0.7131
2 $\exp\; \left( { - \dfrac{ { { {\left\| {{{a}}_1 - {{{a}}_2} } \right\|}^2} } }{ {2{\sigma ^2} } } } \right)$ 0.8520 0.8355 0.7452
3 ${{a}}_1^{\rm{T}}{{a}}_2$ 0.8529 0.8310 0.7472
4 $\gamma {\left( {{{a} }_1^{\rm{T} }{{a} }_2 + 1} \right)^d} + c$ 0.8514 0.8367 0.7544
5 $\exp\; \left( { - \dfrac{ {\left\| {{{a}}_1 - {{{a}}_2} } \right\|} }{ {2{\sigma ^2} } } } \right)\lambda + \left( { {\rm{1 - } }\lambda } \right)\left[ {\gamma { {\left( {{{a}}_1^{\rm{T}}{{a}}_2 + 1} \right)}^d} + c} \right]$ 0.8529 0.8368 0.7554
6 $\exp \;\left( { - \dfrac{ { { {\left\| {{{a}}_1 - {{{a}}_2} } \right\|}^2} } }{ {2{\sigma ^2} } } } \right)\lambda + \left( { {\rm{1 - } }\lambda } \right)\left[ {\gamma { {\left( {{{a}}_1^{\rm{T}}{{a}}_2 + 1} \right)}^d} + c} \right]$ 0.8503 0.8387 0.7555
Tab.3 Comparison of performance of different kernel function
算法模型 数据集1 数据集2 数据集3
MSE RMSE MAPE MSE RMSE MAPE MSE RMSE MAPE
PSO-SVM 564.06 23.75 0.1624 251.54 15.86 0.1724 48.86 6.99 0.2701
LSTM 493.13 22.21 0.1435 234.37 15.31 0.1689 82.88 9.10 0.2702
GRU 495.68 22.26 0.1432 233.978 15.29 0.1676 83.10 9.12 0.2733
CNN-LSTM 483.18 21.92 0.1438 235.89 15.36 0.1663 82.43 9.08 0.2659
CNN-GRU[26] 487.46 22.07 0.1429 229.62 15.15 0.1648 82.44 9.08 0.2640
Bi-LSTM[27] 481.34 21.94 0.1545 223.13 14.94 0.1811 81.58 9.03 0.2703
GA-CKRVM[28] 433.53 20.82 0.1412 161.56 12.71 0.1616 62.24 7.76 0.2347
CNN-Bi-LSTM 477.35 21.84 0.1396 227.11 15.07 0.1622 81.79 9.04 0.2578
SPGAPSO-CKRVM 392.43 19.81 0.1383 161.1 12.69 0.1589 41.09 6.41 0.2232
Tab.4 Comparison of traffic flow predicted by different algorithms and models
Fig.6 Traffic flow prediction results in three stations by SPGAPSO-CKRVM
Fig.7 Traffic flow prediction results in three stations after increasing test set
算法模型 数据集1 数据集2 数据集3
MSE RMSE MAPE MSE RMSE MAPE MSE RMSE MAPE
PSO-SVM 6.0762 2.4650 0.0738 7.2539 2.6933 0.0768 7.1888 2.6812 0.0754
LSTM 5.1940 2.2790 0.0596 5.6122 2.3690 0.0625 5.5984 2.3661 0.0629
GRU 5.1532 2.2701 0.0583 5.6074 2.3681 0.0628 5.6164 2.3699 0.0632
CNN-LSTM 5.1726 2.2740 0.0582 5.3024 2.3027 0.0613 5.5418 2.3541 0.0622
CNN-GRU 5.0687 2.2514 0.0576 5.2964 2.3014 0.0612 5.5469 2.3552 0.0621
Bi-LSTM 5.1557 2.2706 0.0579 5.4228 2.3287 0.0627 5.5691 2.3599 0.0625
GA-CKRVM 4.8118 2.1936 0.0581 5.0895 2.2560 0.0603 5.5145 2.3483 0.0597
CNN-Bi-LSTM 5.0794 2.2538 0.0572 5.1734 2.2745 0.0599 5.5286 2.3513 0.0603
SPGAPSO-CKRVM 4.6656 2.1601 0.0570 4.9809 2.2318 0.0586 5.4214 2.3284 0.0589
Tab.5 Comparison of speed predicted by different algorithms and models
Fig.8 Training time under different circumstances
Fig.9 Result of speedup under different circumstances
Fig.10 Result of road condition prediction
[1]   林浩, 李雷孝, 王慧 支持向量机在智能交通系统中的研究应用综述[J]. 计算机科学与探索, 2020, 14 (6): 901- 917
LIN Hao, LI Lei-xiao, WANG Hui Survey on research and application of support vector machines in intelligent transportation system[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14 (6): 901- 917
doi: 10.3778/j.issn.1673-9418.2001029
[2]   ZHENG C, LI L. The improvement of the forecasting model of short-term traffic flow based on wavelet and ARMA[C]// Proceedings of 2010 8th International Conference on Supply Chain Management and Information Systems. Hong Kong: IEEE, 2011: 1-4.
[3]   TAN M C, WONG S C, XU M C, et al An aggregation approach to short-term traffic flow prediction[J]. Intelligent Transportation Systems, 2009, 10 (1): 60- 69
[4]   CAO J, XU G H, HOU L, et al Detection and estimation for the traffic flow based on Kalman filter[J]. Journal of Beijing Institute of Technology, 2011, 20 (5): 271- 275
[5]   CAO C T, XU J M. Short-term traffic flow predication based on PSO-SVM[C]// 1st International Conference on Transportation Engineering. Chengdu: ASCE, 2007.
[6]   温峻峰, 李鑫, 张浪文 基于粒子群优化的支持向量回归车道饱和度预测[J]. 自动化仪表, 2019, 40 (8): 38- 42
WEN Jun-feng, LI Xin, ZHANG Lang-wen Traffic lane saturation prediction with the support vector regression based on particle swarm optimization[J]. Process Automation Instrumentation, 2019, 40 (8): 38- 42
[7]   FU R, ZHANG Z, LI L. Using LSTM and GRU neural network methods for traffic flow prediction[C]// 31st Youth Academic Annual Conference of Chinese Association of Automation. Wuhan: IEEE, 2016.
[8]   LIU Q, WANG B, ZHU Y Short-term traffic speed forecasting based on attention convolutional neural network for arterials[J]. Computer-Aided Civil and Infrastructure Engineering, 2018, 33 (11): 999- 1016
doi: 10.1111/mice.12417
[9]   LIU Y, ZHENG H, FENG X, et al. Short-term traffic flow prediction with Conv-LSTM[C]// 9th International Conference on Wireless Communications and Signal Processing. Nanjing: IEEE, 2017.
[10]   XIN J, XIAO F H Financial assets price prediction based on relevance vector machine with genetic algorithm[J]. Journal of Convergence Information Technology, 2012, 7 (5): 90- 96
doi: 10.4156/jcit.vol7.issue5.12
[11]   SHEN Z, WANG W, SHEN Q, et al Hybrid CSA optimization with seasonal RVR in traffic flow forecasting[J]. KSII Transactions on Internet and Information Systems, 2017, 11 (10): 4887- 4907
[12]   SHEN Z G, WANG W L, SHEN Q, et al A novel learning method for multi-intersections aware traffic flow forecasting[J]. Neurocomputing, 2020, 398 (7): 477- 484
[13]   王璐媛, 于雷, 孙建平, 等 交通运行指数的研究与应用综述[J]. 交通信息与安全, 2016, 34 (3): 1- 9
WANG Lu-yuan, YU Lei, SUN Jian-ping, et al An overview of studies and applications on traffic performance index[J]. Journal of Transport Information and Safety, 2016, 34 (3): 1- 9
doi: 10.3963/j.issn1674-4861.2016.03.001
[14]   VAZIRI M Development of highway congestion index with fuzzy set models[J]. Transportation Research Record: Journal of the Transportation Research Board, 2002, 1802 (1): 16- 22
doi: 10.3141/1802-03
[15]   KONG X, YANG J, YANG Z. Measuring traffic congestion with taxi GPS data and travel time index[C]// 15th COTA International Conference of Transportation Professionals. Beijing: ASCE, 2015.
[16]   LEVINSON H S, LOMAX T J Developing a travel time congestion index[J]. Transportation Research Record Journal of Transportation Research Board, 1996, 1564 (1): 1- 10
doi: 10.1177/0361198196156400101
[17]   宋顶利, 张昕, 于复兴 并行优化KNN算法的交通运输路况预测模型[J]. 科技通报, 2016, 32 (9): 182- 186
SONG Ding-li, ZHANG Xin, YU Fu-xing Forecasting model of road transportation based on parallel optimized KNN algorithm[J]. Bulletin of Science and Technology, 2016, 32 (9): 182- 186
doi: 10.3969/j.issn.1001-7119.2016.09.040
[18]   晏雨婵, 白璘, 武奇生, 等 基于多指标模糊综合评价的交通拥堵预测与评估[J]. 计算机应用研究, 2019, 36 (12): 3697- 3700
YAN Yu-chan, BAI Lin, WU Qi-sheng, et al Traffic congestion prediction and assessment based on multi-index fuzzy comprehensive evaluation[J]. Application Research of Computers, 2019, 36 (12): 3697- 3700
[19]   TIPPING M E Sparse Bayesian learning and the relevance vector machine[J]. Journal of Machine Learning Research, 2001, 1 (3): 211- 244
[20]   王靖, 张金锁 综合评价中确定权重向量的几种方法比较[J]. 河北工业大学学报, 2001, 30 (2): 52- 57
WANG Jing, ZHANG Jin-suo Comparing several methods of assuring weight vector in synthetical evaluation[J]. Journal of Hebei University of Technology, 2001, 30 (2): 52- 57
doi: 10.3969/j.issn.1007-2373.2001.02.012
[21]   SONG Z, GUO Y, WU Y, et al Short-term traffic speed prediction under different data collection time intervals using a SARIMA-SDGM hybrid prediction model[J]. PLOS ONE, 2019, 14 (6): 1- 19
[22]   DONG E, ZHOU K, TONG J, et al A novel hybrid kernel function relevance vector machine for multi-task motor imagery EEG classification[J]. Biomedical Signal Processing and Control, 2020, 60 (7): 1- 12
[23]   张鹏翔, 刘利民, 马志强 基于MapReduce的层叠分组并行SVM算法研究[J]. 计算机应用与软件, 2015, 32 (3): 172- 176
ZHANG Peng-xiang, LIU Li-min, MA Zhi-qiang Research on cascade-grouping parallel SVM algorithm based on mapreduce[J]. Computer Applications and Software, 2015, 32 (3): 172- 176
doi: 10.3969/j.issn.1000-386x.2015.03.040
[24]   刘泽燊, 潘志松 基于Spark的并行SVM算法研究[J]. 计算机科学, 2016, 43 (5): 238- 242
LIU Ze-shen, PAN Zhi-song Research on parallel SVM algorithm based on Spark[J]. Computer Science, 2016, 43 (5): 238- 242
[25]   ZAHARIA M, CHOWDHURY M, FRANKLIN M J, et al. Spark: cluster computing with working sets[C]// Proceeding of the 2nd USENIX Conference on Hot Topics in Cloud Computing. Berkeley: USENIX Association, 2010: 1-10.
[26]   DU S D, LI T R, GONG X, et al A hybrid method for traffic flow forecasting using multimodal deep learning[J]. International Journal of Computational Intelligence Systems, 2020, 13 (2): 85- 97
[27]   温惠英, 张东冉 基于Bi-LSTM模型的高速公路交通量预测[J]. 公路工程, 2019, 44 (6): 51- 56
WEN Hui-ying, ZHANG Dong-ran Highway traffic volume prediction based on Bi-LSTM model[J]. Highway Engineering, 2019, 44 (6): 51- 56
[28]   邴其春, 龚勃文, 杨兆升, 等 一种组合核相关向量机的短时交通流局域预测方法[J]. 哈尔滨工业大学学报, 2017, 49 (3): 144- 149
BING Qi-chun, GONG Bo-wen, YANG Zhao-sheng, et al A short-term traffic flow local prediction method of combined kernel function relevance vector machine[J]. Journal of Harbin Institute of Technology, 2017, 49 (3): 144- 149
doi: 10.11918/j.issn.0367-6234.2017.03.023
[1] Zhen-hong MA,Zhen LIU,Sheng-yong YIN,Rong-wei MA,Ke-ping YAN. Experimental study on melanoma cell ablation by high-voltage nanosecond pulsed electric field[J]. Journal of ZheJiang University (Engineering Science), 2021, 55(6): 1168-1174.
[2] Ping YANG,Dan WANG,Zi-jian KAGN,Tong LI,Li-hua FU,Yue-ren YU. Prediction model of paroxysmal atrial fibrillation based on pattern recognition and ensemble CNN-LSTM[J]. Journal of ZheJiang University (Engineering Science), 2020, 54(5): 1039-1048.
[3] XIONG Shu-sheng, REN Xiao-shuai, XIE Lian, LI Wei, WU Xi-jiang. Numerical simulation of HCNG engine on GT-Power[J]. Journal of ZheJiang University (Engineering Science), 2013, 47(2): 273-279.
[4] OU Li-Yong, DU Shu-Xin. Multi-agent based coordination of public detection resources[J]. Journal of ZheJiang University (Engineering Science), 2010, 44(1): 81-86+123.