近年來，全球溫室氣體大量排放使全世界平均溫度不斷升高，進而導致極端氣候之頻率不斷上升，衍生自然災害增加日趨嚴重例如旱災、颶風、地震及最常發生的洪水等災害；然而，臺灣地小人稠，位於太平洋西側亞熱帶地區與板塊交界處，時常面臨颱風與地震的威脅，堪稱是天然災害的高風險地區。
全球極端氣候變遷影響，發生短延時強降雨、暴潮及滿潮等影響頻率恐將漸趨頻繁，造成局部低漥地區排水不易形成積淹水之情事可能會比以往都要來的嚴峻。因此，本研究提出透過純文字型態之社群媒體所發布之資訊，以文字探勘及時間序列分析技術為基礎，蒐集本市近年來颱風豪雨事件期間所造成之淹水災情資料，總計292篇文章與12,484則推文，經由詞頻分析評估出積淹水災情之代表性關鍵詞，透過衡量指標決定其參數，進行積淹水災情通報預警之指標，進而建立預警機制，提升積淹水災情通報作業時間，提供相關防救災人員參考。
本研究透過5場事件案例，依相同的時間間隔條件下，不同滑動視窗大小等參數，進行不同門檻值之災情路段預警衡量指標評估，實驗結果發現，在相同的時間間隔條件下，滑動視窗大小為90分鐘且門檻值為20時，整體F-measure值(F=0.315)較高，故採用90分鐘之滑動視窗及門檻值為20等參數，進行災情通報預警之建立；本研究以2016年梅姬颱風事件為驗證，以臺南市安平路(運河旁)為例，因適逢暴雨及滿潮時刻(中央氣象局將軍潮位站，最高潮位約發生於09月27日20時)，以致運河無法順利排入外海，而產生局部溢淹之情形，以該案例之實驗結果顯示可提早約1小時30分鐘之時間，提供相關防救災人員與相關防救災系統參考與應用，以利發揮資源與人力之調度。

Recently, ceaseless rising of global average temperature leads to extreme climates which derive natural disasters turn to be increasingly serious, such as drought, hurricane, earthquake and flood, etc. However, Taiwan is a small densely populated area, located in subtropical region of western Pacific and in the area of plates connection which make it in the high-risk region of natural disaster with the threats of typhoon and earthquakes.
The frequency of heavy rains, storm surge and high tide increase sharply which may lead to more serious ponding and water flooding. Recently, the social media becomes an important resource of sharing information. In consequence, this research came up with the idea of collecting flood disasters caused during the period of typhoon and heavy rain days of the city from the plain text messages released by the social media by means of text mining and time series. In this thesis, the experiment collects 292 articles and 12,484 tweets for analysis. The representative key words of flooding disasters were evaluated based on word frequency analysis; the parameters were determined by measuring the indicators; and the indicators for warnings of flooding disasters are also made to further establish warning mechanism, add notification time for flooding disasters as well as providing relevant disaster relief personnel with references.
The research not only predict flood disaster but also find out possible happening places. According to the experimental results, several parameters are suggested what to use. The sliding window of 90 minutes and the sliding gap of 20 minutes will get the best F-measure value (F=0.315). Based on the experimental results of the case, relevant disaster relief and prevention system can be active 1 hour and 30 minutes before the existed system.

目錄
第1章 緒論 1
1.1. 前言 1
1.2. 研究動機與目的 2
1.3. 研究範圍與限制 3
1.4. 研究流程 4
第2章 文獻探討 5
2.1 社群媒體 5
2.2 自然災害預警之研究 8
2.2.1 社群媒體之應用 9
2.2.2 數值模擬與預警系統 11
2.3 文字探勘與滑動視窗 12
2.4 小結 14
第3章 研究方法 17
3.1 研究架構 17
3.2 資料蒐集與處理模組 18
3.3 事件分析模組 21
3.4 積淹水災情預警與路段估計模組 23
3.4.1 積淹水災情預警模組 24
3.4.2 積淹水災情路段估計模組 25
3.5 小結 26
第4章 系統建置與驗證 27
4.1 資料來源 27
4.2 實驗設計與結果 28
4.2.1 評估指標 29
4.2.2 實驗一 ：積淹水通報關鍵詞 30
4.2.3 實驗二：積淹水災情預警通報衡量指標 33
4.2.4 實驗三：積淹水災情路段預警衡量指標 35
4.3 事件實驗驗證 37
第5章 結論以及未來方向 40
5.1 研究成果 40
5.2 未來研究方向 42
參考文獻 44

參考文獻
Chen, J., Hill, A. A., ＆ Urbano, L. D. (2009). A GIS-based model for urban flood inundation. Journal of Hydrology, 373(1-2), 184-192. doi:10.1016/j.jhydrol.2009.04.021
Chu, C. S. J. (1995). Time-series segmentaiotn - a sliding window approach. Information Sciences, 85(1-3), 147-173. doi:10.1016/0020-0255(95)00021-g
Dallachiesa, M., Jacques-Silva, G., Gedik, B., Wu, K. L., ＆ Palpanas, T. (2015). Sliding windows over uncertain data streams. Knowledge and Information Systems, 45(1), 159-190. doi:10.1007/s10115-014-0804-5
de Albuquerque, J. P., Herfort, B., Brenning, A., ＆ Zipf, A. (2015). A geographic approach for combining social media and authoritative data towards identifying useful information for disaster management. International Journal of Geographical Information Science, 29(4), 667-689. doi:10.1080/13658816.2014.996567
Fayyad, U., Piatetsky-Shapiro, G., ＆ Smyth, P. (1996). From data mining to knowledge discovery in databases. AI magazine, 17(3), 37.
Feldman, D., Contreras, S., Karlin, B., Basolo, V., Matthew, R., Sanders, B., . . . Luke, A. (2016). Communicating flood risk: Looking back and forward at traditional and social media outlets. International Journal of Disaster Risk Reduction, 15, 43-51. doi:10.1016/j.ijdrr.2015.12.004
Gao, H., Barbier, G., ＆ Goolsby, R. (2011). Harnessing the crowdsourcing power of social media for disaster Relief. Ieee Intelligent Systems, 26(3), 10-14.
Guan, X. Y., ＆ Chen, C. (2014). Using social media data to understand and assess disasters. Natural Hazards, 74(2), 837-850. doi:10.1007/s11069-014-1217-1
Hsu, C. T., Yeh, K. C., ＆ Yang, J. C. (2000). Depth-averaged two-dimensional curvilinear explicit finite analytic model for open-channel flows. International Journal for Numerical Methods in Fluids, 33(2), 175-202. doi:10.1002/(sici)1097-0363(20000530)33:2(175::aid-fld8)3.0.co;2-8
Hsu, M. H., Chen, S. H., ＆ Chang, T. J. (2000). Inundation simulation for urban drainage basin with storm sewer system. Journal of Hydrology, 234(1-2), 21-37. doi:10.1016/s0022-1694(00)00237-7
Li, H. L. (2015). On-line and dynamic time warping for time series data mining. International Journal of Machine Learning and Cybernetics, 6(1), 145-153. doi:10.1007/s13042-014-0254-0
Losiewicz, P., Oard, D. W., ＆ Kostoff, R. N. (2000). Textual data mining to support science and technology management. Journal of Intelligent Information Systems, 15(2), 99-119. doi:10.1023/a:1008777222412
McCallum, I., Liu, W., See, L., Mechler, R., Keating, A., Hochrainer-Stigler, S., . . . Moorthy, I. (2016). Technologies to support community flood disaster risk reduction. International Journal of Disaster Risk Science, 7(2), 198-204. doi:10.1007/s13753-016-0086-5
Middleton, S. E., Middleton, L., ＆ Modafferi, S. (2014). Real-time crisis mapping of natural disasters using social media. Ieee Intelligent Systems, 29(2), 9-17. doi:10.1109/mis.2013.126
Newsam, S. (2010). Crowdsourcing what is where: Community-contributed photos as volunteered geographic information. Ieee Multimedia, 17(4), 36-45. doi:10.1109/mmul.2010.79
Pohl, D., Bouchachia, A., ＆ Hellwagner, H. (2016). Online indexing and clustering of social media data for emergency management. Neurocomputing, 172, 168-179. doi:10.1016/j.neucom.2015.01.084
Ryang, H., ＆ Yun, U. (2016). High utility pattern mining over data streams with sliding window technique. Expert Systems with Applications, 57, 214-231. doi:10.1016/j.eswa.2016.03.001
Sakaki, T., Okazaki, M., ＆ Matsuo, Y. (2013). Tweet analysis for real-time event detection and earthquake reporting system development. Ieee Transactions on Knowledge and Data Engineering, 25(4), 919-931. doi:10.1109/tkde.2012.29
Salton, G., ＆ Buckley, C. (1988). Term-weighting approaches in automatic text retrieval. Information Processing ＆ Management, 24(5), 513-523. doi:10.1016/0306-4573(88)90021-0
Takahashi, B., Tandoc, E. C., ＆ Carmichael, C. (2015). Communicating on Twitter during a disaster: An analysis of tweets during Typhoon Haiyan in the Philippines. Computers in Human Behavior, 50, 392-398. doi:10.1016/j.chb.2015.04.020
Tang, Z. H., Zhang, L. G., Xu, F. H., ＆ Vo, H. (2015). Examining the role of social media in California's drought risk management in 2014. Natural Hazards, 79(1), 171-193. doi:10.1007/s11069-015-1835-2
Tsai, T. L., Chen, H. E., ＆ Yang, J. C. (2008). Numerical modeling of rainstorm-induced shallow landslides in saturated and unsaturated soils. Environmental Geology, 55(6), 1269-1277. doi:10.1007/s00254-007-1075-1
Tsai, T. L., ＆ Yang, J. C. (2006). Modeling of rainfall-triggered shallow landslide. Environmental Geology, 50(4), 525-534. doi:10.1007/s00254-006-0229-x
Yun, U., ＆ Lee, G. (2016). Sliding window based weighted erasable stream pattern mining for stream data applications. Future Generation Computer Systems-the International Journal of Escience, 59, 1-20. doi:10.1016/j.future.2015.12.012
Chow, V.T. (1964). Runoff. In: Chow, V.T. (Ed.), Handbook of Applied Hydrology.McGraw-Hill, New York, pp. 14-1–14-54.
Cunge, J. A., Holly, F. M., and Verwey, A. (1980). Practical Aspects of Computational River Hydraulics, Pitman Publishing Ltd., London.
維基百科. (2016). 批踢踢-維基百科，自由的百科全書. from https://zh.wikipedia.org/wiki/%E6%89%B9%E8%B8%A2%E8%B8%A2