建築物火警初期派遣之關聯性研究-以新北市為例

A Study of Relationships in Building Fire Preliminary Dispatch for Fire Department of New Taipei City

林韋伶(Wei-Ling Lin)

指導教授 ： 方鄒昭聰

國立臺北大學/商學院/資訊管理研究所/碩士(2012年)

摘要

台灣過去在消防火警派遣領域中，針對派遣類別的律定是經由相關消防專家開會討論而產生，並未有文獻以過去發生的火警案件資料為基礎，以資料分析角度，利用量化資料分析報案資訊以及派遣車種之間的關係,原因為過去之資料未以結構化方式進行儲存整理，因此未能進行探討。本研究目的在利用資料分析的角度，針對建築物火警初期派遣行為進行報案資訊與是否派遣該車種之間關連性的研究，並針對資料分析的結果與專家訂定的派遣原則進行比較分析其異同之處。研究蒐集2009年至2011年間建築物火警之案例資料，首先進行結構化處理，接著針對處理過後的540筆案件資料進行資料預處理，最後進行派遣行為分析。分析可分三大部分，第一部分先了解案件資料結構特徵以次序統計進行分析，第二部分以羅吉斯迴歸進行分析，探討單一車種的是否派遣行為，可由何種類別變數解釋。第三部分則使用多元尺度分析法進行分析，將派遣行為相似的車種歸為一類集合，目的在發現實際派遣行為上基本車組集合、稀少車組集合以及其他車組集合。結果發現可解釋車種是否派遣行為的派遣類別實際分析結果與專家訂定的結果有異。另外發現水箱車與救護車為建築物火警上會派出的基本車種，且行成基本派遣車組；排煙車種與滅火機車為稀少派遣車種，其餘之車種則為功能車種，用以解決火警案件中某一情況的派遣類別而額外選擇搭配的車種。研究以車種是否派遣的角度討論派遣類別與車種之間的關係，並提出將半結構資料轉換為結構化資料進而得以進行統計分析的研究流程設計，是將質性資料轉為量化資料的探索性研究。在實務上，幫助組織實際了解內部派遣行為與專家訂定規則之異同，並可於後續之派遣系統設計或派遣規則律定上，皆可透過量化的方式進行衡量與參考。

關鍵字

建築物火警初期派遣 ； 消防派遣 ； 救災救護指揮中心

並列摘要

In fire preliminary dispatch, dispatchers need to gather some dispatch categories information through fire calls until the arrival of fire service. The dispatch categories information will influence individual fire engine to be dispatched. In the past of Taiwan, these dispatch categories were often generated from expert's through relevant meeting, and there had no any literatures applying data analysis on history records to find dispatch behavior between categories and fire engines. The main reason cause above discussion was that past fire case data was stored in the forms of unstructured data, which is difficult for fire department to analysis. The purpose of this study is applying data analysis approach and quantified data to analysis relationships between dispatch categories and the dispatch behavior of whether dispatch fire engine, then compare the similarities and differences between the result of data analysis and the dispatch principle expert set. The study used 2009~2011,totally 3 years fire case data from the New Taipei City, first structured the case data and retained 540 cases data to do data preprocessing, then analysis the behavior of fire dispatch. The analysis can be divided into three parts, the first part aim to understand data structure characteristics of the case, using the order statistics to analyze. Second part is to investigate the behavior about what kind of categorical variables can explain whether to send a type of fire engine, and this was analyzed by logistic regression. The third part use multidimensional scaling analysis to analyze, vehicles have similar dispatch behavior will be considered as the same group, the purpose of this part is to discover basic dispatch fire engine set, rare dispatch fire engine set and other dispatch fire engine set in actual dispatch behaviors. Our research found that the category which can explain the vehicle's dispatch behavior was different from the ruled which generated from expert. Besides, the basic dispatch fire engine sets are water tank and ambulance, while the rare dispatch fire engine sets are fume exhauster truck and fire motorcycle. In addition, others are functional engine set in order to resolve situations in the fire case, and were dispatched with basic dispatch fire engine set. This research is an exploratory study discusses relationships between categories and fire engine in whether to dispatch the fire engine perspective or not. The research designed a way to convert semi-structure dispatch case data into structured data, transformed qualitative data into quantitative data. In practical contribution, the study can help fire department understand internal dispatch behavior and compare the similarities or differences with experts rule set. Furthermore, our research provide a different aspect of analysis to help fire department building fire dispatch system and setting dispatch rule.

並列關鍵字

Building Fire Preliminary Dispatch ； Fire dispatch ； Dispatch center

參考文獻

[1] E.Turban., R. Sandra. & D. Delen. (2010). Decision support system and business intelligence systems. (9th ed.): Prentice Hall.

[2] Kruskal, J. B. (1964). Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29(1): 1-27.

[2] 伊相志. (2008). SQL Server 2008 Data Mining 資料採礦.台北: 悅知文化.

英文部分

[3] Hosmer, D. & Lemeshow, S. (2000). Applied Logistic Regression (2th ed.). New York: John Wiley & Sons, Inc.