鐵路運輸是在陸上運輸系統中肇事率最小的運輸系統，但往往每當發生事故輕者阻礙交通、重者車毀人亡，其中又以鐵路與公路二大運輸系統之平面交會處的平交道最易發生事故，而平交道事故也最常帶來極大生命及財產損失，因此本研究其目的在於建立一套鐵路平交道事故之分析模式可以用來預測未來可能發生肇事情形，也可以藉此了解鐵路平交道事故之相關因素。本研究透過民國87 年台灣省政府交通處出版一套四冊之平交道改善規劃，包括當時交通事故資料、平交道種類、道路幾何型態、列車班次數與平均每年每日交通量等曝光量資料，並藉由台灣鐵路局運務處部門民國91年至96年行車事故資料，整理成各處平交道事故資料，最後透過卜瓦松迴歸、負二項迴歸、類神經倒傳遞網路，三種模式分別來構建平交道事故次數、死亡人數、受傷人數三種預測模式，並找出最適模式及相關顯著影響因素以供未來相關鐵路單位做為未來肇事預測。
研究結果發現，事故次數結果顯示以卜瓦松模式與負二項模式預測皆優於類神經倒傳遞網路預測結果；影響平交道相關模式的因素中，以全日列車班次為最重要影響因子。

In all kind of the land transportation, the railway transportation system has the lowest rate of accident. While once an accident is happened, it usually causes life and property loss. Especially the level crossing which is crossed between railway and highway. Slight accident will tie up traffic, and serious accident will cause death and car crash. This research forecasts the possiblity of accident happening by setting a railway crossing accident analysis model. Based on the research, we can realize more about the correlation factor. The research uses data from Crossing Improvement Plan published by Taiwan Provincial Government in 1998 as references. It consists of both historical accident data and railway level crossing related data in Taiwan, such as crossing types, highway geometric characteristics, daily trains and average annual daily traffic (AADT), etc. We try to reorganize the traffic accident data since 2002 to 2007, collected from Transportation Department, Taiwan Railway Administration, to the level crossing accident data. Finally, the research could be applied to Poisson Regression Model, Negative Binomial regression, and Back-Propagation Neural Network for evaluating the prediction of the level crossing accident frequency models, fatalities models, and hurt models then to find the fittest model with the major factor which is relative significantly. It can provide the relation with the TRA for forecasting accident in the future.
As a consequence of this research, we can find Poisson Regression Model and Negative Binomial regression are better than Back-Propagation Neural Network. The most significant factor of effect on level crossing model is the number of pass trains.

一、中文部分
1.JJ. Scott Long著，鄭旭智等譯 (2002)，「類別與受限依變項的迴歸統計模式」，pp.120-134，弘智文化。
2.劉應興編譯 (2003)，「類別資料分析導論」，華泰文化。
3.劉應興編譯 (1998)，「非線性迴歸與相關分析」，華泰書局。
4.葉怡成 (2003)，「類神經網路模式應用與實作」，第八版，儒林書局。
5.台灣省政府交通處 (1998、1999)，「台灣地區鐵路平交道改善規劃 (一) ~ (四)」。
6.交通部 (1996)，「鐵路立體交叉及平交道防護設施設置標準與費用分攤規則」。
7.交通部運輸研究所 (2008)，「建立台鐵安全系統績效指標之研究（期末報告初稿）」。
8.交通部台灣鐵路局 (2007)，96年鐵路年鑑。
9.林錦田、莊振昌 (1999)，「平交道改善專案報告」，台鐵資料季刊，第302期，頁1-49。
10.林豐福、周永暉 (1996)，「台灣地區鐵路平交道事故之研究」，交通部運輸研究所。
11.林豐福、張開國、吳熙仁 (1999)，「以科技儀器提昇平交道防護功能之研究」，台鐵資料季刊，第302期，頁50-63。
12.施伯杰 (2008)，「以故障樹與事件樹分析法探討平交道事故風險」，碩士論文，國立成功大學交通管理科學研究所。
13.施鈞明 (2004)，「應用系統模擬於平交道之風險分析」，碩士論文，國立成功大學交通管理科學研究所。
14.唐鵬州 (2006)，「平交道模擬模式構建與肇事風險分析」，碩士論文，國立成功大學交通管理科學研究所。
15.陳火庸 (2006)，「台鐵行車事故肇事因子之研究」，碩士論文，逢甲大學交通工程與管理學系碩士在職專班。
16.黃維崧 (2006)，「影響台鐵平交道事故因素之研究」，碩士論文，國立交通大學運輸科技與管理學系。
17.楊凱評 (2007)，「平交道事故頻率與衝擊分析」，碩士論文，國立成功大學交通管理科學研究所。
18.詹坤益 (2002)，「先進平交道安全控制蓄意闖越行為之系統設計」，碩士論文，國立成功大學交通管理科學研究所。
19.蔡孟紋 (2006)，「平交道事故分析之研究」，碩士論文，國立成功大學交通管理科學研究所。
20.鄭子良 (2002)，「先進偵測系統對鐵路平交道安全之影響研究」，碩士論文，國立中央大學土木工程學系。
21.駱思斌 (2002)，「鐵路事故嚴重程度之研究」，碩士論文，國立成功大學交通管理科學研究所。
二、英文部分
1.U.S. Department of Transportation, (1986), “Railroad-Highway Grade Crossing Handbook,” Second Edition, FHWA-TS-86-215.
2.Austin, R.D. and Carson, J.L. (2002), “An Alternative Accident Prediction Model for Highway-Rail Interfaces,” Accident Analysis and Prevention, vol. 34, pp. 31–42.
3.Chang, L.Y., (2005),“Analysis of Freeway Accident Frequencies:Negative Binomial Regression Versus Artificial Neural Network,” Safety Science, vol. 43, No.8, pp. 541-557.
4.Lee, C.K. and Hu, S.R. (2007), “Accident Risk at a Railway Level Crossing,” Proceedings of the Eastern Asia Society for Transportation Studies, vol. 6.
5.Lee, J., Nam, D.,and Park D. (2005), “Analyzing the Relationship between Grade Crossing Elements and Accidents,” Journal of the Eastern Asia Society for Transportation Studies, vol. 6, pp. 3658 – 3668.
6.Lewis,C.D.(1982), Industrial and Business Forecasting Methods,Southampton：The Camelot Press Ltd.
7.Martin,C.A. and Witt,S.F, (1989), “Accuracy of Econometric Forecasts of tourism,” Annals of Tourism Research, vol. 16, pp. 407-428.
8.McCullagh, P., Nelder, J. A.,(1989),“Generalized Linear Models,” Chapman and Hall.
9.McFadden, D., Train, K. and Typ, W. E.(1977), “An Application of Diagnostic Test For Independence Form Irrelevant Alternative Property of the Mutinimial Logit Model,” Transportation Research , vol. 637, pp. 39-46.
10.Oh, J., Washington, S.P.,and Nam, D. (2006), “Accident Prediction Model for Railway-highway Interfaces,” Accident Analysis and Prevention, vol. 38, pp. 346–356.
11.Park, Y.J. and Saccomanno F.F. (2005), “Evaluating Factors Affecting Safety at Highway-Railway Grade Crossings,” Transportation Research Record, No. 1918, pp. 1-9.
12.Raub, R.A., (2006),“Examination of Highway-Rail Grade Crossing Collisions Over 10 Years in Seven Midwestern States,” ITE Journal, vol. 76, No. 4, pp. 21-26.
13.Saccomanno, F.F. and Lai X.M., (2005), “A Model for Evaluating Countermeasures at Highway-Railway Grade Crossings,” Transportation Research Record, No. 1918, pp. 18-25 .