Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90526
Title: | 應用社會網絡分析於跨境走私犯罪風險評估—以臺灣海上走私犯罪為例 Applying Social Network Analysis to Assess Transnational Trafficking Risk: The Case of Maritime Trafficking in Taiwan |
Authors: | 張伊君 Yi-Chun Chang |
Advisor: | 曹承礎 Seng-cho T. Chou |
Keyword: | 海上走私犯罪,風險管理,社會網絡分析,梯度提升決策樹,圖神經網絡分析, Maritime trafficking,risk management,social network analysis,gradient boost decision tree,GNN, |
Publication Year : | 2023 |
Degree: | 博士 |
Abstract: | 海上走私犯罪指的是在海上非法貿易或是載運違禁品入境,包括毒品、菸酒等經濟貨物、野生動物、人口販運等,針對運輸途徑斷絕通路,可以有效打擊走私犯罪。臺灣四面環海,最常見的走私模式就是透過漁船進行海上走私,目前臺灣查獲海上走私犯罪的情資來源主要都是透過線報,一般檢查查獲的走私犯罪案件數量非常少,若能提供安檢人員更多有關風險評估的資訊,可增加海巡機關針對船舶風險管理的效率。
研究主題分成三部分,首先建構海上走私網絡(Maritime trafficking network)並識別走私集團中的重要角色,第二部分則是提出使用SNA作為特徵值進行風險評估的「海上走私風險評估模型(Maritime Trafficking Risk Assessment,MTRA)」,使用社會網絡分析(SNA)法結合梯度提升決策樹(Gradient Boosting Decision Tree,GBDT)方法進行風險評估,由過去的海上活動行為預測未來可能會有走私犯罪風險的人員,最後使用圖神經網絡(Graph Neural Network, GNN)方法進行風險偵測,並比較圖形資料中不同關係權重對於分析結果的影響,找出適合的風險評估方式。在MTRA模型中加入風險因子與SNA因子,研究過程使用16種方始來處理資料不平衡的問題並選擇出最佳的方案,經實驗結果顯示SNA因子之後能夠能讓模型整體準確率上升20%以上,而且MTRA模型能發現60%的高風險者。採取學術單位與實務界一起共同合作模式,研究成果可提供執法機關未來進行漁船風險管理系統參考。 Maritime trafficking crimes involve the illegal trade or importation of contraband by sea, including goods such as tobacco, alcohol, wild animals, drugs, as wellas human trafficking. Taiwan is surrounded by the sea, and smuggling commonly occurs via vessels. Cutting off transportation can effectively combat smuggling. The successful interception of maritime trafficking crimes primarily stems from information provided from imformants but rarely from vessel inspections. Therefore, it is important to provide more information to law enforcement to improve the efficiency of inspectors and risk management. In this paper, there are three major tasks. First, to establish a maritime trafficking networks (MTN), and identify the key members in smuggling groups. Second, to propose a framework of risk assessment model called “Maritime trafficking risk assessment“ (MTRA) model . The MTRA model combines social network analysis (SNA) and gradient boosting decision tree (GDBT). It is a binary classification model to identify the high risk members. Finally, the current research uses a graph neural network (GNN) with different definitions for comparison of the performance of risk assessment. Besides, to deal with the imbalanced problems, it takes 16 different methods to process imbalence data and choose the best solution. The experment results show that the accuracy and F1-score of MTRA model rose more than 20 % after adding the feature of SNA factors. This research cooperates with the Coast Guard Agency in Taiwan, and the results could be a pototype to establish a vessel risk management system in the future. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90526 |
DOI: | 10.6342/NTU202301119 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 資訊管理學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-111-2.pdf Restricted Access | 3.98 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.