Publications-Theses
Article View/Open
Publication Export
-
題名 信用卡盜刷模型偵測：分別以類神經網路及支援向量機之模型成效比較
Credit card fraud model detection：Comparison of the model effectiveness of neural network and support vector machine作者 陳宇慈 貢獻者 蔡炎龍<br>周冠男
陳宇慈關鍵詞 深度學習
信用卡盜刷
類神經網路
異常偵測
資料不平衡
資料降維
Deep Learning
Credit Card Fraud
Neural Network
Anomaly Detection
Data Imbalance
Dimension Reduction日期 2021 上傳時間 1-Jul-2021 21:38:01 (UTC+8) 摘要 在現今的時代中，網路購物、線上購物已為消費者進行購物的主要管道。而在付款方式的選擇上，「信用卡支付」又相較於「超商取貨付款」，多了更多的便利性。加上銀行業者、以及電商業者時常會提供信用卡付款優惠當成誘因，吸引消費者使用信用卡付款。但是，任何事物有正面也有反面，而信用卡所帶來的便利的背後即是盜刷的風險。目前信用卡盜刷主要可以分成三種盜刷形式，依序分別為偽冒申請、盜刷、偽卡交易，依照台灣財團法人聯合信用卡中心統計數字顯示，2018年信用卡盜刷金額高達23.59億元。而在研究流程的部分，為了研究模型準確度的保證，本研究先透過合成少數類過取樣技術 (SMOTE) 演算法將原始資料集進行資料類別不平衡的處理，接著將已處理好的數據透過全連結神經網路進行信用卡盜刷模型的建立，另外，也以三種方法進行降維模型的設計，三種方法分別是使用主成分分析 (Principal components analysis, PCA)、全連結神經網路(NN)、以及函數式API (Function API)，最後在模型成效評估時則以支援向量機（Support Vector Machine）作為分類模型的建置，最後則是透過混合矩陣來評估模型分類的效果。而從實證結果中我們可以發現，以類神經網絡來建立信用卡盜刷模型的模型準確度表現是最好的。
In this era, online shopping has become the main channels for consumers to shop. When it comes to the choice of payment methods, "Credit Card " is more convenient than "convenient store Pickup and Payment". In addition, banks and e-commerce companies often offer credit card payment discounts as an incentive to attract consumers to pay with credit cards. However, everything has its pros and cons, and behind the convenience brought by credit cards is the risk of fraud. At present, credit card fraud can be divided into three types of fraudulent use, which are counterfeit applications, fraudulent use, and counterfeit card transactions. According to statistics from the Taiwan Consortium Credit Card Center, the amount of fraudulent use of credit cards in 2018 was as high as 2.359 billion NTD.In the research process, in order to ensure the accuracy of the research model, in this research, first uses the Synthetic Minority Oversampling Technology (SMOTE) algorithm to deal with the data imbalance, and then the processed data is passed through the fully connected neural network, which is used to build the credit card fraud model. In addition, three methods are used to build the dimensionality reduction model. The three methods are the use of principal components analysis (PCA) and the fully connected neural network (NN). And functional API (Function API). In the end of the research process, support vector machine (Support Vector Machine) is used as the construction of the classification model in the evaluation of the model effect. The effect of the model classification is evaluated through the mixing matrix. From the final results, we can find that the accuracy of the model that uses a neural network to build a credit card fraud model is the best.參考文獻 Agarwal, A., El-Ghazawi, T., El-Askary, H., & Le-Moigne, J. (2007, December 1). Efficient Hierarchical-PCA Dimension Reduction for Hyperspectral Imagery. https://doi.org/10.1109/ISSPIT.2007.4458191ÇAVDAR, İ., & FARYAD, V. (2019). New Design of a Supervised Energy Disaggregation Model Based on the Deep Neural Network for a Smart Grid. Energies, 12(7), 1217. https://doi.org/10.3390/en12071217Chen, C., Leu, J., & Prakosa, S. W. (2018, April 1). Using autoencoder to facilitate information retention for data dimension reduction. https://doi.org/10.1109/IGBSG.2018.8393545陳昇瑋 文字作者 Chenshengwei, & Wen, Y. (2019). 人工智慧在台灣 : 產業轉型的契機與挑戰 = AI Taiwan / Ren gong zhi hui zai tai wan : chan ye zhuan xing de qi ji yu tiao zhan = AI Taiwan. 天下雜誌股份有限公司, 大和圖書有限公司 Tai Bei Shi: Tian Xia Za Zhi Gu Fen You Xian Gong Si, [Xin Bei Shi.Ghosh, & Reilly. (1994, January 1). Credit card fraud detection with a neural-network. https://doi.org/10.1109/HICSS.1994.323314Hu, C., Hou, X., & Lu, Y. (2014). Improving the Architecture of an Autoencoder for Dimension Reduction. 2014 IEEE 11th Intl Conf on Ubiquitous Intelligence and Computing and 2014 IEEE 11th Intl Conf on Autonomic and Trusted Computing and 2014 IEEE 14th Intl Conf on Scalable Computing and Communications and Its Associated Workshops. https://doi.org/10.1109/uic-atc-scalcom.2014.50Ibrahim, M. F. I., & Al-Jumaily, A. A. (2016, December 1). PCA indexing based feature learning and feature selection. https://doi.org/10.1109/CIBEC.2016.7836122Jain, V., Agrawal, M., & Kumar, A. (2020, June 1). Performance Analysis of Machine Learning Algorithms in Credit Cards Fraud Detection. https://doi.org/10.1109/ICRITO48877.2020.9197762Kazemi, Z., & Zarrabi, H. (2017, December 1). Using deep networks for fraud detection in the credit card transactions. https://doi.org/10.1109/KBEI.2017.8324876Khatri, S., Arora, A., & Agrawal, A. P. (2020, January 1). Supervised Machine Learning Algorithms for Credit Card Fraud Detection: A Comparison. https://doi.org/10.1109/Confluence47617.2020.9057851Li, J., Fong, S., & Zhuang, Y. (2015, December 1). Optimizing SMOTE by Metaheuristics with Neural Network and Decision Tree. https://doi.org/10.1109/ISCBI.2015.12Malini, N., & Pushpa, M. (2017). Analysis on credit card fraud identification techniques based on KNN and outlier detection. 2017 Third International Conference on Advances in Electrical, Electronics, Information, Communication and Bio-Informatics (AEEICB). https://doi.org/10.1109/aeeicb.2017.7972424Mittal, S., & Tyagi, S. (2019, January 1). Performance Evaluation of Machine Learning Algorithms for Credit Card Fraud Detection. https://doi.org/10.1109/CONFLUENCE.2019.8776925Srivastava, A., Yadav, M., Basu, S., Salunkhe, S., & Shabad, M. (2016, March 1). Credit card fraud detection at merchant side using neural networks. Retrieved October 23, 2020, from IEEE Xplore website: https://ieeexplore.ieee.org/document/7724348/塚本邦尊, 文字作者 Bangzun Zhongben, Dianyi Shantian, Wenxiao Daze, & Yongyu Zhuang. (2020). 東京大學資料科學家養成全書 : 使用Python動手學習資料分析 = 東京大学のデータサイエンティスト育成講座 : 東京大学のデータサイエンティスト育成講座 / Dong jing da xue zi liao ke xue jia yang cheng quan shu : shi yongPython dong shou xue xi zi liao fen xi = dong jing da xue no de - ta sa i e nn te i su to yu cheng jiang zuo : dong jing da xue no de - ta sa i e nn te i su to yu cheng jiang zuo. 臉譜出版: 英屬蓋曼群島商家庭傳媒股份有限公司城邦分公司發行, Tai Bei Shi. 描述 碩士
國立政治大學
企業管理研究所(MBA學位學程)
108363092資料來源 http://thesis.lib.nccu.edu.tw/record/#G0108363092 資料類型 thesis dc.contributor.advisor 蔡炎龍<br>周冠男 zh_TW dc.contributor.author (Authors) 陳宇慈 zh_TW dc.creator (作者) 陳宇慈 zh_TW dc.date (日期) 2021 en_US dc.date.accessioned 1-Jul-2021 21:38:01 (UTC+8) - dc.date.available 1-Jul-2021 21:38:01 (UTC+8) - dc.date.issued (上傳時間) 1-Jul-2021 21:38:01 (UTC+8) - dc.identifier (Other Identifiers) G0108363092 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/136030 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 企業管理研究所(MBA學位學程) zh_TW dc.description (描述) 108363092 zh_TW dc.description.abstract (摘要) 在現今的時代中，網路購物、線上購物已為消費者進行購物的主要管道。而在付款方式的選擇上，「信用卡支付」又相較於「超商取貨付款」，多了更多的便利性。加上銀行業者、以及電商業者時常會提供信用卡付款優惠當成誘因，吸引消費者使用信用卡付款。但是，任何事物有正面也有反面，而信用卡所帶來的便利的背後即是盜刷的風險。目前信用卡盜刷主要可以分成三種盜刷形式，依序分別為偽冒申請、盜刷、偽卡交易，依照台灣財團法人聯合信用卡中心統計數字顯示，2018年信用卡盜刷金額高達23.59億元。而在研究流程的部分，為了研究模型準確度的保證，本研究先透過合成少數類過取樣技術 (SMOTE) 演算法將原始資料集進行資料類別不平衡的處理，接著將已處理好的數據透過全連結神經網路進行信用卡盜刷模型的建立，另外，也以三種方法進行降維模型的設計，三種方法分別是使用主成分分析 (Principal components analysis, PCA)、全連結神經網路(NN)、以及函數式API (Function API)，最後在模型成效評估時則以支援向量機（Support Vector Machine）作為分類模型的建置，最後則是透過混合矩陣來評估模型分類的效果。而從實證結果中我們可以發現，以類神經網絡來建立信用卡盜刷模型的模型準確度表現是最好的。 zh_TW dc.description.abstract (摘要) In this era, online shopping has become the main channels for consumers to shop. When it comes to the choice of payment methods, "Credit Card " is more convenient than "convenient store Pickup and Payment". In addition, banks and e-commerce companies often offer credit card payment discounts as an incentive to attract consumers to pay with credit cards. However, everything has its pros and cons, and behind the convenience brought by credit cards is the risk of fraud. At present, credit card fraud can be divided into three types of fraudulent use, which are counterfeit applications, fraudulent use, and counterfeit card transactions. According to statistics from the Taiwan Consortium Credit Card Center, the amount of fraudulent use of credit cards in 2018 was as high as 2.359 billion NTD.In the research process, in order to ensure the accuracy of the research model, in this research, first uses the Synthetic Minority Oversampling Technology (SMOTE) algorithm to deal with the data imbalance, and then the processed data is passed through the fully connected neural network, which is used to build the credit card fraud model. In addition, three methods are used to build the dimensionality reduction model. The three methods are the use of principal components analysis (PCA) and the fully connected neural network (NN). And functional API (Function API). In the end of the research process, support vector machine (Support Vector Machine) is used as the construction of the classification model in the evaluation of the model effect. The effect of the model classification is evaluated through the mixing matrix. From the final results, we can find that the accuracy of the model that uses a neural network to build a credit card fraud model is the best. en_US dc.description.tableofcontents 第一章 緒論 8第一節 研究背景與動機 8第二節 信用卡盜刷定義 9第三節 研究目的 10第四節 研究架構 10第二章 文獻探討 12第一節 機器學習(machine learning)與深度學習(Deep learning)定義 12第二節 機器學習預測信用卡盜刷應用 13第三節 深度學習預測信用卡盜刷應用 14第四節 降維(Dimension reduction/Embedding) 14第五節 主成分分析 15第六節 函數式API (Functional API) 15第三章 研究方法 18第一節 研究資料分析 18第二節 探索式資料分析（Exploratory Data Analysis，簡稱EDA) 18第三節 特徵工程與特徵標準化 21第四節 資料不平衡處理 21第六節 模型建立 27第七節 模型預測能力衡量指標 34第四章 實證結果 37第一節 研究工具 37第二節 切分資料訓練集與測試集 37第三節 模型建置設定 37第四節 PCA模型實證結果(PCA+SVM) 39第五節 類神經網路模型實證結果(SMOTE +SVM)41第六節 Functional API模型實證結果(Functional API +SVM) 43第七節 模型實證結果之綜合比較 44第五章 結論與建議 46第一節 研究結論 46第二節 研究限制與未來建議 46第六章 參考文獻 48 zh_TW dc.format.extent 4696082 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0108363092 en_US dc.subject (關鍵詞) 深度學習 zh_TW dc.subject (關鍵詞) 信用卡盜刷 zh_TW dc.subject (關鍵詞) 類神經網路 zh_TW dc.subject (關鍵詞) 異常偵測 zh_TW dc.subject (關鍵詞) 資料不平衡 zh_TW dc.subject (關鍵詞) 資料降維 zh_TW dc.subject (關鍵詞) Deep Learning en_US dc.subject (關鍵詞) Credit Card Fraud en_US dc.subject (關鍵詞) Neural Network en_US dc.subject (關鍵詞) Anomaly Detection en_US dc.subject (關鍵詞) Data Imbalance en_US dc.subject (關鍵詞) Dimension Reduction en_US dc.title (題名) 信用卡盜刷模型偵測：分別以類神經網路及支援向量機之模型成效比較 zh_TW dc.title (題名) Credit card fraud model detection：Comparison of the model effectiveness of neural network and support vector machine en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) Agarwal, A., El-Ghazawi, T., El-Askary, H., & Le-Moigne, J. (2007, December 1). Efficient Hierarchical-PCA Dimension Reduction for Hyperspectral Imagery. https://doi.org/10.1109/ISSPIT.2007.4458191ÇAVDAR, İ., & FARYAD, V. (2019). New Design of a Supervised Energy Disaggregation Model Based on the Deep Neural Network for a Smart Grid. Energies, 12(7), 1217. https://doi.org/10.3390/en12071217Chen, C., Leu, J., & Prakosa, S. W. (2018, April 1). Using autoencoder to facilitate information retention for data dimension reduction. https://doi.org/10.1109/IGBSG.2018.8393545陳昇瑋 文字作者 Chenshengwei, & Wen, Y. (2019). 人工智慧在台灣 : 產業轉型的契機與挑戰 = AI Taiwan / Ren gong zhi hui zai tai wan : chan ye zhuan xing de qi ji yu tiao zhan = AI Taiwan. 天下雜誌股份有限公司, 大和圖書有限公司 Tai Bei Shi: Tian Xia Za Zhi Gu Fen You Xian Gong Si, [Xin Bei Shi.Ghosh, & Reilly. (1994, January 1). Credit card fraud detection with a neural-network. https://doi.org/10.1109/HICSS.1994.323314Hu, C., Hou, X., & Lu, Y. (2014). Improving the Architecture of an Autoencoder for Dimension Reduction. 2014 IEEE 11th Intl Conf on Ubiquitous Intelligence and Computing and 2014 IEEE 11th Intl Conf on Autonomic and Trusted Computing and 2014 IEEE 14th Intl Conf on Scalable Computing and Communications and Its Associated Workshops. https://doi.org/10.1109/uic-atc-scalcom.2014.50Ibrahim, M. F. I., & Al-Jumaily, A. A. (2016, December 1). PCA indexing based feature learning and feature selection. https://doi.org/10.1109/CIBEC.2016.7836122Jain, V., Agrawal, M., & Kumar, A. (2020, June 1). Performance Analysis of Machine Learning Algorithms in Credit Cards Fraud Detection. https://doi.org/10.1109/ICRITO48877.2020.9197762Kazemi, Z., & Zarrabi, H. (2017, December 1). Using deep networks for fraud detection in the credit card transactions. https://doi.org/10.1109/KBEI.2017.8324876Khatri, S., Arora, A., & Agrawal, A. P. (2020, January 1). Supervised Machine Learning Algorithms for Credit Card Fraud Detection: A Comparison. https://doi.org/10.1109/Confluence47617.2020.9057851Li, J., Fong, S., & Zhuang, Y. (2015, December 1). Optimizing SMOTE by Metaheuristics with Neural Network and Decision Tree. https://doi.org/10.1109/ISCBI.2015.12Malini, N., & Pushpa, M. (2017). Analysis on credit card fraud identification techniques based on KNN and outlier detection. 2017 Third International Conference on Advances in Electrical, Electronics, Information, Communication and Bio-Informatics (AEEICB). https://doi.org/10.1109/aeeicb.2017.7972424Mittal, S., & Tyagi, S. (2019, January 1). Performance Evaluation of Machine Learning Algorithms for Credit Card Fraud Detection. https://doi.org/10.1109/CONFLUENCE.2019.8776925Srivastava, A., Yadav, M., Basu, S., Salunkhe, S., & Shabad, M. (2016, March 1). Credit card fraud detection at merchant side using neural networks. Retrieved October 23, 2020, from IEEE Xplore website: https://ieeexplore.ieee.org/document/7724348/塚本邦尊, 文字作者 Bangzun Zhongben, Dianyi Shantian, Wenxiao Daze, & Yongyu Zhuang. (2020). 東京大學資料科學家養成全書 : 使用Python動手學習資料分析 = 東京大学のデータサイエンティスト育成講座 : 東京大学のデータサイエンティスト育成講座 / Dong jing da xue zi liao ke xue jia yang cheng quan shu : shi yongPython dong shou xue xi zi liao fen xi = dong jing da xue no de - ta sa i e nn te i su to yu cheng jiang zuo : dong jing da xue no de - ta sa i e nn te i su to yu cheng jiang zuo. 臉譜出版: 英屬蓋曼群島商家庭傳媒股份有限公司城邦分公司發行, Tai Bei Shi. zh_TW dc.identifier.doi (DOI) 10.6814/NCCU202100493 en_US
