基于PCA和HMM的汽车保有量预测方法

孙璐; 郁烨; 顾文钧

doi:10.19818/j.cnki.1671-1637.2013.02.014

基于PCA和HMM的汽车保有量预测方法

doi: 10.19818/j.cnki.1671-1637.2013.02.014

孙璐^{1, 2,},
郁烨¹,
顾文钧²

1.
东南大学交通学院, 江苏南京 210096
2.
美国天主教大学土木工程系, 华盛顿特区 20064

基金项目:

国家自然科学基金项目 51010044

国家自然科学基金项目 U1134206

详细信息

作者简介:
孙璐(1972-), 男, 上海人, 东南大学教授, 长江学者, 工学博士, 从事道路交通安全研究

中图分类号: U491.14
计量
- 文章访问数: 673
- HTML全文浏览量: 95
- PDF下载量: 1330
- 被引次数: 0
出版历程
- 收稿日期: 2012-09-17
- 刊出日期: 2013-04-25

Car ownership prediction method based on principal component analysis and hidden Markov model

SUN Lu^{1, 2
,},
YU Ye¹,
GU Wen-jun²

1.
School of Transportation, Southeast University, Nanjing 210096, Jiangsu, China
2.
Department of Civil Engineering, The Catholic University of America, Washington DC 20064, USA

More Information

Author Bio:
SUN Lu (1972-), male, professor, PhD, +86-25-83792619, sunl@cua.edu

摘要

摘要: 分析了常用的汽车保有量预测方法, 提出了一种新的基于主成分分析和隐马尔可夫模型的汽车保有量预测方法。选取国民总收入、人均GDP、人口总数量、城市化率、固定资产投资总额、进出口总额、城镇居民人均可支配收入、钢材产量、公路货运量、公路客运量、社会消费品零售总额11个指标作为汽车保有量的主要影响因素, 运用主成分分析提取了主要影响因素的主成分。以提取的主成分与汽车保有量分别作为自变量、因变量, 建立了回归分析模型。以汽车保有量回归预测值的年增长率为隐状态, 以回归预测值与实际值的相对误差为可见信号, 建立了隐马尔科夫模型, 并对的汽车保有量回归预测值进行修正。分析结果表明: 基于1994~2008年的中国汽车保有量及其主要影响因素的历史数据, 应用提出的方法得到2009、2010年的汽车保有量修正值分别为6.220 96×10⁷、7.825 12×10⁷ veh; 与2009、2010年实际汽车保有量比较, 相对误差分别为-0.95%、0.30%。可见, 基于主成分分析和隐马尔科夫模型的汽车保有量预测方法具有良好的预测精度, 能够适用于短期预测。
- 汽车保有量预测 /
- 隐马尔可夫模型 /
- 主成分分析 /
- 回归分析 /
- 灰色预测
Abstract: The usual prediction methods of car ownership were analyzed, a new car ownership prediction method based on principal component analysis(PCA) and hidden Markov model(HMM) was put out. The 11 indexes including gross national income, per capita GDP, total population number, urbanization rate, total fixed asset investment, gross import and export, urban resident disposable income, steel output, highway passenger transport volume, highway freight transport volume, total retail sales of consumer goods were taken as the main influence factors of car ownership, and PCA was used to extract the principal components of main influence factors. The principal component and car ownership were taken as independent variable and dependent variable respectively, and the regression analysis model was set up. The annual growth rates of regression prediction values for car ownership were taken as hidden state, the relative errors between regression prediction values and actual values were taken as visible signal, the hidden Markov model was built, and the regression prediction values of car ownership were modified.Analysis result shows that based on car ownerships and the historical data of main influence factors in 1994-2008, the numbers of modified car ownership in 2009 and 2010 are 6.220 96×10⁷ and 7.825 12×10⁷ by using the proposed method.Compared with the actual values of car ownership in 2009 and 2010, relative errors are-0.95% and 0.30% respectively. So car ownership prediction method based on PCA and HMM has a high prediction accuracy and is suitable for short-term prediction.
- car ownership prediction /
- hidden Markov model /
- principal component analysis /
- regression analysis /
- grey prediction

HTML全文

图 1 汽车保有量与主成分的关系

Figure 1. Relationship between car ownership and principal component

下载: 全尺寸图片幻灯片

表 1 主要影响因素

Table 1. Main influence factors

下载: 导出CSV

表 2 相关系数

Table 2. Correlation coefficients

下载: 导出CSV

表 3 方差贡献率

Table 3. Total variance explained

下载: 导出CSV

表 4 汽车保有量的回归预测值

Table 4. Regression prediction values of car ownership

下载: 导出CSV

表 5 分类标准

Table 5. Classification standard

下载: 导出CSV

表 6 状态转移序列和可见信号序列

Table 6. State transfer sequence and visible signal sequence

下载: 导出CSV

表 7 状态转移概率

Table 7. State transfer probability

下载: 导出CSV

表 8 可见信号的分布概率

Table 8. Distribution probability of visible signals

下载: 导出CSV

表 9 预测结果

Table 9. Prediction result

下载: 导出CSV

参考文献(21)

[1]	张益, 陈淑燕, 王炜. 短时交通量时间序列智能复合预测方法概述[J]. 公路交通科技, 2006, 23(8): 139-142. doi: 10.3969/j.issn.1002-0268.2006.08.032 ZHANG Yi, CHEN Shu-yan, WANG Wei. Survey of traffic volume time series intelligent compound forecasting methods[J]. Journal of Highway and Transportation Research and Development, 2006, 23(8): 139-142. (in Chinese). doi: 10.3969/j.issn.1002-0268.2006.08.032
[2]	POTOGLOU D, SUSILO Y. Comparison of car ownership models[J]. Transportation Research Record, 2008(2076): 97-105.
[3]	陈远通. 汽车保有量预测技术方法及其应用[D]. 广州: 华南理工大学, 2010. CHEN Yuan-tong. The methodology and implementation of car ownerships forecast technique[D]. Guangzhou: SouthChina University of Technology, 2010. (in Chinese).
[4]	DARGAY J, GATELY D. Income's effect on car and vehicle ownership, worldwide: 1960-2015[J]. Transportation Research Part A: Policy and Practice, 1999, 33(2): 101-138. doi: 10.1016/S0965-8564(98)00026-3
[5]	仲伟周, 王军. 基于因子分析法的中国民用汽车保有量影响因素分析及其政策含义[J]. 统计与信息论坛, 2008, 23(11): 16-20. doi: 10.3969/j.issn.1007-3116.2008.11.003 ZHONG Wei-zhou, WANG Jun. The analysis of the effectivefactor of China's civilian vehicle holdings and its policy implications based on the factor analysis[J]. Statistics and Information Forum, 2008, 23(11): 16-20. (in Chinese). doi: 10.3969/j.issn.1007-3116.2008.11.003
[6]	杨艳妮, 陆化普. 基于支持向量机的城市私人汽车保有量滚动预测[J]. 公路工程, 2010, 35(5): 40-43. https://www.cnki.com.cn/Article/CJFDTOTAL-ZNGL201005010.htm YANG Yan-ni, LU Hua-pu. An improved prediction modelfor the stock of urban private cars based on support vectormachine[J]. Highway Engineering, 2010, 35(5): 40-43. (inChinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZNGL201005010.htm
[7]	曹晓飞, 邵春福. 运用弹性系数法预测北京机动车保有量[J]. 道路交通与安全, 2008, 8(4): 31-34. https://www.cnki.com.cn/Article/CJFDTOTAL-DLJA200804010.htm CAO Xiao-fei, SHAO Chun-fu. Forecast of vehicle amountby elastic coefficient method in Beijing[J]. Road Traffic and Safety, 2008, 8(4): 31-34. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-DLJA200804010.htm
[8]	何明, 过秀成. BP神经网络主成分分析法在汽车保有量预测中的应用[J]. 交通与计算机, 2007, 25(4): 96-99. https://www.cnki.com.cn/Article/CJFDTOTAL-JTJS200704026.htm HE Ming, GUO Xiu-cheng. Application of BP neural network principal component analysis to forecasting the automobile population[J]. Computer and Communications, 2007, 25(4): 96-99. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-JTJS200704026.htm
[9]	周骞, 杨东援. 基于多相关因素的汽车保有量预测神经网络方法[J]. 公路交通科技, 2001, 18(6): 71-74. https://www.cnki.com.cn/Article/CJFDTOTAL-GLJK200106032.htm ZHOU Qian, YANG Dong-yuan. Multi-relative factor forecast of vehicle population by neural network[J]. Journal of Highway and Transportation Research and Development, 2001, 18(6): 71-74. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-GLJK200106032.htm
[10]	朱海清, 陈学武. 路网服务水平影响下的机动车规模预测[J]. 交通运输工程与信息学报, 2005, 3(4): 80-85. doi: 10.3969/j.issn.1672-4747.2005.04.016 ZHU Hai-qing, CHEN Xue-wu. Amount prediction of motor-driven vehicles under the influences of road network service level[J]. Journal of Transportation Engineering and Information, 2005, 3(4): 80-85. (in Chinese). doi: 10.3969/j.issn.1672-4747.2005.04.016
[11]	朱开永, 周圣武, 娄可元, 等. 基于私家车保有量预测与调控的灰色模型研究[J]. 中国矿业大学学报, 2008, 37(6): 868-872. doi: 10.3321/j.issn:1000-1964.2008.06.026 ZHU Kai-yong, ZHOU Sheng-wu, LOU Ke-yuan, et al. Research on grey model about forecast and control for total number of private cars[J]. Journal of China University of Mining and Technology, 2008, 37(6): 868-872. (in Chinese). doi: 10.3321/j.issn:1000-1964.2008.06.026
[12]	朱灿, 周和平, 钟璧樯. 基于混合RBF网络模型的汽车保有量组合预测[J]. 长沙理工大学学报: 自然科学版, 2011, 8(2): 13-16. doi: 10.3969/j.issn.1672-9331.2011.02.003 ZHU Can, ZHOU He-ping, ZHONG Bi-qiang. Forecastingon vehicle population based on combined model with mixedneural network[J]. Journal of Changsha University ofScience and Technology: Natural Science, 2011, 8(2): 13-16. (in Chinese). doi: 10.3969/j.issn.1672-9331.2011.02.003
[13]	王琦, 王花兰. 基于熵值法的城市汽车保有量组合预测[J]. 交通科技与经济, 2009, 11(6): 53-55. https://www.cnki.com.cn/Article/CJFDTOTAL-KJJJ200906017.htm WANG Qi, WANG Hua-lan. Combination forecasting of vehiclepopulation based on the entropy method[J]. Technology andEconomy in Areas of Communications, 2009, 11(6): 53-55. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-KJJJ200906017.htm
[14]	亓芳芳. 基于面板数据模型的中国民用汽车消费需求预测研究[D]. 合肥: 中国科学技术大学, 2010. QI Fang-fang. Forecasting research on the consumptiondemand of civilian vehicle in China based on panel datamodel[D]. Hefei: University of Science and Technology ofChina, 2010. (in Chinese).
[15]	吴明, 王莹莹, 冯琪. 基于主成分分析的综合运输通道布局模型研究[J]. 公路交通科技, 2011, 28(1): 154-158. https://www.cnki.com.cn/Article/CJFDTOTAL-GLJK201101031.htm WU Ming, WANG Ying-ying, FENG Qi. Research on com-prehensive transportation channel layout model based on maincomponent analysis[J]. Journal of Highway and Transporta-tion Research and Development, 2011, 28(1): 154-158. (inChinese). https://www.cnki.com.cn/Article/CJFDTOTAL-GLJK201101031.htm
[16]	张文峰, 史忠科. 主成分分析法在城市道路交通安全评价中的应用[J]. 计算机工程与应用, 2007, 43(36): 246-248. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG200736075.htm ZHANG Wen-feng, SHI Zhong-ke. Application of principalcomponent analysis in evaluation of city road traffic safety[J]. Computer Engineering and Applications, 2007, 43(36): 246-248. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG200736075.htm
[17]	RABINER L R. A tutorial on hidden Markov models andselected applications in speech recognition[J]. Proceedings ofthe IEEE, 1989, 77(2): 257-286.
[18]	RABINER L R, JUANG B H. An introduction to hiddenMarkov models[J]. IEEE ASSP Magazine, 1986, 3(1): 4-16.
[19]	张庆生, 齐勇, 侯迪, 等. 基于隐马尔科夫模型的上下文感知活动计算[J]. 西安交通大学学报, 2006, 40(4): 398-401. https://www.cnki.com.cn/Article/CJFDTOTAL-XAJT200604006.htm ZHANG Qing-sheng, QI Yong, HOU Di, et al. Context-aware activity computing based on hidden Markov model[J]. Journal of Xi'an Jiaotong University, 2006, 40(4): 398-401. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-XAJT200604006.htm
[20]	吕颖钊, 贺拴海. 在役桥梁承载力模糊可靠性的马尔科夫预测[J]. 长安大学学报: 自然科学版, 2005, 25(4): 39-43. https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200504010.htm LU Ying-zhao, HE Shuan-hai. Prediction of Markov fuzzyreliability for existing bridge[J]. Journal of Chang'an University: Natural Science Edition, 2005, 25(4): 39-43. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200504010.htm
[21]	张珍. 基于隐马尔科夫模型的波动率预测[D]. 广州: 华南理工大学, 2012. ZHANG Zhen. Volatility forecastion via hidden Markovmodels[D]. Guangzhou: South China University of Technology, 2012. (in Chinese).