English  |  正體中文  |  简体中文  |  全文筆數/總筆數 : 55613/89938 (62%)
造訪人次 : 11054070      線上人數 : 290
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU Library IR team.
搜尋範圍 查詢小技巧:
  • 您可在西文檢索詞彙前後加上"雙引號",以獲取較精準的檢索結果
  • 若欲以作者姓名搜尋,建議至進階搜尋限定作者欄位,可獲得較完整資料
  • 進階搜尋
    請使用永久網址來引用或連結此文件: http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/33834

    題名: 交叉路口兩車肇事鑑定預測模式之研究
    其他題名: The study on predict model of two vehicles accident authentication at intersection
    作者: 林文賢;Lin, Wen-hsien
    貢獻者: 淡江大學運輸管理學系碩士班
    范俊海;Fan, Chun-hai
    關鍵詞: 交通事故;肇事鑑定;層級分析法;類神經網路;預測模式;判中率;Traffic Accident Accident Authentication;Analytic Hierarchy Process ( AHP );Artificial Neural Network ( ANN );Predict mode;hit ratio
    日期: 2007
    上傳時間: 2010-01-11 04:35:16 (UTC+8)
    摘要: 當交通事故( Traffic Accident )發生時,因需保持肇事現場狀況,所以無法立即將車輛移動;或因警察處理交通事故時間過久,進而影響交通流量( Traffic Flow )。再者各區車輛車輛行車事故鑑定委員會由法律、車輛工程與交通工程三方面專家所組成,而每位鑑定委員對於法律認知不同、解讀見解不一或跡證不足而對肇事鑑定( Accident Authentication ) 的結論產生爭議及審件速度緩慢等問題。所以站在鑑定委員的立場上,若有較一致的肇事鑑定影響因素判斷順序之規則,可協助其進一步做肇事原因判定的工作,如此除能減少鑑定委員之間的爭議外,也能加速審查案件。此對於申請肇事鑑定的當事人與對造人,可以最快的速度判斷事故的肇事原因而節省寶貴的時間,是一個值得研究的課題。
    最後本研究透過所構建的類神經網路預測模式與AHP預測模式做一整體的驗證與比較,結果顯示藉由客觀的類神經網路預測模式之判中率72.53%優於透過全省各區車輛車輛行車事故鑑定委員主觀意見所構建的AHP預測模式之判中率35.16% ,即表示類神經網路預測模式仍是目前較好的模式。另外運用類神經網路中平均絕對誤差率( MAPE )得知,以兩個隱藏層,在第一層有12個單元和第二層有7個單元時其MAPE值為19.12%,績效指標屬於良好的評估中可了解本預測模式符合分析交叉路口兩車肇事鑑定影響因素之判別準則。
    Since we have to maintain the traffic accident scene while Traffic Accident occurs or the police officer takes a lot of time dealing with the traffic accident, the traffic flow is certainly being influenced by accidents.
    The Traffic Accident Appraisal Committee ( TAAC ) is composed of experts of three fields: experts on law, vehicle engineering and traffic engineering.
    Members of TAAC will have quarrels on the conclusions of accident authentication due to different comprehension to the laws quoted and the insufficient evidences. To the perspective on the commissioners of TAAC, the identical criterions on priorities of the factors on accident authentication is necessary. It can not only reduce the quarrels of the commissioners but also speed up the investigates. And more over, it can assist commissioners to conclude the cause of the accident. Since the predict model can determine the cause of accidents fast and accurate , saving litigants’ valuable time, it is obviously a classic issue for us to dig into.
    To the perspectives above, the study on building the predict model and determining the weight of vehicle accident authentication effective factors is necessary. Taking analysis the priorities of vehicle accident authentication effective factors as this study’s main idea, through the 273 sample data collected from TAAC of Taipei country which conform with the definition of two vehicles accident at intersection, listing the 18 variables that may effect the judgments on the cause of accident authentication by literature review. Using discriminant analysis to conduct 12 variables that effect the accident authentication. Using the information above, designing an AHP questionnaire that contains three aspects and eleven criteria. By the subjective opinions of commissions from different TAAC of Taiwan filled in the questionnaire, we can analysis the weights and the priority of each criteria and get to know the importance of effective variables of two vehicle accident authentication at intersection.
    This study compared the two predict models conduct by neural network and AHP model. The result showed that the Neural network predict model has the accurate of 72.53%, which is much better than the AHP predict model with the accurate of 35.16%. The conclusion showed that the Neural network predict model is so far a better model on this subject.
    Using MAPE of neural network method we got MAPE=19.12% while the first layer with 12 units and the second layer with 7 units. The evaluation result with good preference shows that the predict model conform the criteria of two vehicles accident authentication effective factors.
    顯示於類別:[運輸管理學系暨研究所] 學位論文


    檔案 大小格式瀏覽次數



    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - 回饋