Research on Seasonal Index Based on Dynamic Clustering of the Daily Railway Passenger Flow Title

Abstract:

The railway passenger flow is greatly impacted by different months and weeks of the season, and the impact is periodic. Accurate evaluation of the seasonal index for predicting the railway passenger flow is of key importance. Based on this background, the paper proposes an algorithm for calculating the seasonal index which is impacted by both months and weeks. The railway passenger flow between different OD(Origination Destination) is affected by months and weeks quite different. Therefore the paper focuses on the method for effective calculation of the month index and week index on the basis of time series clustering. When adopting hierarchical cluster, general Euclidean distance and its expansion used as a similarity metric is widely applied in time series comparison, however, this distance measurement is not robust enough for the processed data. Dynamic time warping is a pattern matching algorithm based on nonlinear dynamic programming technique. It is applied to calculate month and week index to get seasonal index that defined in this paper, which has good application value for predicting the passenger flow.

• Periodical: Advanced Materials Research (Volume 981)
• Pages: 966-971
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.981.966
• Online since: July 2014
• Authors: Zheng Zheng Wei*, Fu Zhang Wang
• Keywords: Clustering, Dynamic Time Warping Distance, Passenger Flow Forecast, Seasonal Index

References

[1] Philidep, Hanse Francis. Translated by Feng Jianqiang. Time sequence model in business and economical prediction. [M]. Beijing. China Renmin University Press. 2002. 12.

[2] Chang Linxing. Research on railway passenger flow based on seasonal index. Railway transportation and ecomomic. 2006. 7.

[3] Pan Xiaowei. Multiple regression and time series analysis in application of passenger flow prediction in the expo. Economic research.

[4] Weng Yingjun, Zhu Zhongying. Time-series data clustering algorithm based on dynamic time warping. computer simulation. 2004. 3.

[5] Lv Qingji．Two kinds of time series comparison of outlier mining method. Journal of statistics, 2005. 1.

[6] Feng Hongwei．The research and application of data mining technology. Ph.D. Dissertation, northwestern polytechnical university. 2004. 7.

[7] Wang Jianzhou, Yang Yong. Outlier data identification and mining based on the statistical distribution of the wavelet analysis of time series. Journal of northwest normal university (natural science edition). 2004. 2.

[8] He Lin, Wu Lingda, Cai Yinchao. Clustering algorithm in data mining. Computer application research. 2007. Volumn 1.

[9] R O Duda, P E Hart, D G Stork. Pattern Classification(2nd Edition) [M]. New York: Wiley, (2001).