基於社群網路資料之捷運運量預測

MRT Demand Prediction through Social Media

林泉亨(Chuan-Heng Lin)

指導教授 ： 陳柏華

國立臺灣大學/工學院/土木工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.11261

摘要

隨著社群網路的興起和行動裝置之普及，無論人們在社群網路上打卡，發文，或是上傳照片，都可能成為瞭解人群移動的依據。近幾年來，社群網路人數的大量增加使其成為一熱門之研究議題，而其研究課題大多圍繞者旅遊推薦，和使用者之間的關係，皆屬於個人與個人之間的課題研究，較少探討群體或是應用在系統上面的問題，因此本研究旨在將社群網路資料應用在大眾運輸系統運量需求預測之可行性。透過文字探勘方法建立話題特徵及影像偵測方法取得社群網路 - Instagram 中影像資料中人臉的的特徵，將其結合機器學習方法中的支持向量機(Support Vector Machine)，隨機森林(Random Forest)，及隨機梯度提升方法(Stochastic Gradient Boosting)，建立預測短期各捷運站的旅客出站數之模式。本研究利用政府公開的捷運站每日各站資料，作為驗證的依據。驗證結果顯示在本研究對社群網路所提取之特徵中文字特徵具有較好之結果，且其MAPE值落在良好預測之範圍內。初步成果顯示在本研究所建議之特徵提取架構之下，能有良好預測結果並有潛力應用於實務中。

關鍵字

機器學習 ； 社群網路 ； 捷運 ； 文本探勘 ； 影像偵測

並列摘要

With the technological improvements of mobile devices and the increasing number of social media posts, there are more and more data on human mobility based on which information could potentially be extracted. Current research related to social media are mostly focused on inter-person behaviors. Conversely, related topics on system level performances are rarely discussed. This thesis applies feature extraction methods on quantitative, textual, and image data to retrieve useful features from social media. In addition, a machine learning pipeline based on support vector machine, random forest and stochastic gradient boosting is constructed for a short-term transportation demand forecast. Furthermore, real-world datasets from Instagram together with the demand data of the Taipei Metro Rapid Transit system are demonstrated in this work. Validation results show that social media has the potential to enhance the forecasting accuracy.

並列關鍵字

Social Media ； Topic Models ； Support Vector Machine ； Random Forest ； Stochastic Gradient Boosting ； Transportation

參考文獻

[51] C. D. Lewis. Industrial and business forecasting methods : a practical guide to exponential smoothing and curve fitting / Colin D. Lewis. Butterworth Scientific London, 1982.

[34] Jerome Friedman, Trevor Hastie, and Robert Tibshirani. The elements of statistical learning, volume 1. Springer series in statistics Springer, Berlin, 2001.

[22] Thomas L Griffiths and Mark Steyvers. Finding scientific topics. Proceedings of the National Academy of Sciences, 101(suppl 1):5228–5235, 2004.

[25] Wenyi Zhao, Rama Chellappa, P Jonathon Phillips, and Azriel Rosenfeld. Face recognition: A literature survey. ACM computing surveys (CSUR), 35(4):399–458, 2003.

[17] David M Blei, Andrew Y Ng, and Michael I Jordan. Latent dirichlet allocation. the Journal of machine Learning research, 3:993–1022, 2003.