A robust AI-based automated trading system

强大的基于人工智能的自动交易系统

• 批准号: 556396-2020
• 负责人: Hemmati, Hadi
• 金额: $ 3.97万
• 依托单位: University of Calgary
• 依托单位国家: 加拿大
• 项目类别: Alliance Grants
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=706321
• 关键词: robust; AI; based; automated; trading

The use of AI platforms within the financial markets industry of trading is still new. Hence this industry has a lot of room to grow. Day trading strategies can easily be manipulated by a trader's emotions which leads to poor decisions on both beginners and professional traders. Therefore, there is a need for a platform that predicts the market patterns and suggests the entry and exit points based on multiple strategies, while helping traders to stay disciplined when the market is highly volatile. One requirement of such solution is to predict the market based on the users mood in online social networks (e.g., by sentiment analysis of twits). However, such AI-enabled automated trading systems must be very robust with respect to noises, malicious behavior, and attacks. Attacks can be done by individual adversaries as well as social bots, making them harder to track down. In addition to AI models robustness, the implementation of these systems also needs to be reliable with respect to any environmental failure, uncertainties, or stress (e.g., network glitches, power shut down, overload transactions, etc.). Therefore, in this project we propose an integrated automated robustness testing for AI-enabled automated trading systems. The solution covers both software-level robustness testing as well as robustness of the AI algorithms, especially when the inputs are sentiments learnt from online social networks such as Twitter.

人工智能平台在金融市场交易行业的使用仍然是新的。因此，这个行业有很大的发展空间。日内交易策略很容易被交易者的情绪所操纵，从而导致新手和专业交易者做出糟糕的决定。因此，需要一个平台来预测市场模式，并根据多种策略建议进入和退出点，同时帮助交易者在市场高度波动时保持纪律。这种解决方案的一个要求是基于在线社交网络中的用户情绪来预测市场（例如，通过对twits的情绪分析）。然而，这种人工智能支持的自动交易系统必须在噪声、恶意行为和攻击方面非常强大。攻击既可以由个人对手进行，也可以由社交机器人进行，这使得它们更难被追踪。除了AI模型的鲁棒性之外，这些系统的实现还需要在任何环境故障、不确定性或压力（例如，网络故障、电源关闭、过载事务等）方面是可靠的。因此，在这个项目中，我们提出了一个集成的自动化鲁棒性测试，用于支持人工智能的自动交易系统。该解决方案既涵盖了软件层面的稳健性测试，也涵盖了人工智能算法的稳健性，尤其是当输入是从Twitter等在线社交网络学习到的情绪时。