Trade Finance Fraud Detection Project in Dual Use Goods with Machine Learning and Visual Analytics

利用机器学习和视觉分析的双重用途商品贸易融资欺诈检测项目

• 批准号: 104413
• 金额: $ 48.13万
• 依托单位: TRAYDSTREAM LIMITED
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2018
• 资助国家: 英国
• 起止时间: 2018 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=104413
• 关键词: Trade; Finance; Fraud; Detection; Project

Traydstream has developed three core technology solutions that are set to disrupt global trade finance. A Trade Finance document review by an expert typically takes between 2-3 hours or as much as between 10 and 45 days between the four involved parties i.e. a buyer, a seller and a bank representing each of these.Traydstream has reduced (currently) a single review to less than 20 minutes. If all four parties used the Traydstream Platform concurrently, then such a saving, even if rounded up to 1 hour and calculated against a 10 day timeframe, would amount to 99.6% reduction or almost 100 times faster.Traydstream's disruptive ambitions do not end there. The company now wishes to advance its technology by developing a cutting edge software solution in collaboration with Swansea University to prove that it is possible to detect trade finance fraud using human interpretable machine learning and visual analytics in the complex area of Dual Use Goods (DUGs). Current manual paper based practice is resource consuming and often inconsistent. According to the Royal United Services Institute (RUSI) there is a real threat to international security that proliferators of trade finance are able to obtain the component parts of highly regulated and controlled items such as biological or chemical weapons, in pursuit of war or terrorist activities. This project brings together our experience in trade finance and Swansea's expertise in deep learning and visualisation to tackle a complex issue, particularly in face of the trade and regulatory uncertainties associated with Brexit.A successful outcome from this project has a real potential to disrupt this global practice and place the region and the UK at the forefront of trade finance fraud detection and prevention.

Traffic Stream开发了三种核心技术解决方案，将颠覆全球贸易融资。由专家进行的贸易融资文件审查通常需要2-3小时或多达10至45天的时间，涉及的四个当事方，即买方，卖方和代表每一方的银行。Traffic Stream（目前）将单次审查减少到不到20分钟。如果四方同时使用Traffic Stream平台，那么这样的节省，即使四舍五入到1小时，以10天的时间计算，也相当于减少99.6%或快近100倍。Traffic Stream的颠覆性野心并没有就此结束。该公司现在希望通过与斯旺西大学合作开发尖端软件解决方案来推进其技术，以证明在复杂的两用品（DUG）领域使用人类可解释的机器学习和视觉分析来检测贸易融资欺诈是可能的。目前的手工纸质实践是资源消耗和经常不一致。据皇家联合服务研究所（入寺）有一个真实的威胁，国际安全的贸易融资扩散者能够获得高度监管和控制的项目，如生物或化学武器的组成部分，在追求战争或恐怖活动。该项目汇集了我们在贸易融资方面的经验和斯旺西在深度学习和可视化方面的专业知识，以解决一个复杂的问题，特别是在面对与Brexit相关的贸易和监管不确定性的情况下。该项目的成功结果具有真实的潜力，可以打破这种全球做法，并将该地区和英国置于贸易融资欺诈检测和预防的最前沿。