Financial Risk assessment of AI industry using a new machine Learning model

人工智能业务已经成为最受期待的科技行业，吸引了大量投资。人工智能技术与我们生活的方方面面密切相关，推动了传统行业的创新，包括医疗、教育、电信、制造、零售、金融等，而且在国家层面，它影响着公民的安全和隐私、气候保护、产业治理和经济政策。各国致力于建立和发展长期的人工智能战略。因此，人工智能产业的健康和可持续发展至关重要。然而，有许多人工智能初创企业因遭受财务风险而破产，这给利益相关者和社会带来了不利影响。财务风险存在于企业经营管理的各个环节，受各种不可控因素的影响，可能导致企业财务状况恶化、信用违约甚至破产。建立一个预警模型来评估和预测人工智能公司的财务困境势在必行。金融风险评估本质上是一个多准则决策分析问题，其目的是通过信息聚合对多个备选方案进行排序或选择最佳方案。本研究将使用证据推理（ER）方法，这是一种数据驱动的机器学习方法来预测企业财务风险。与传统行业不同，人工智能行业的运营具有更大的不确定性，比如研发投入更大，技术更新更快，资金回收期、盈利模式或市场需求预测的不确定性更高。单靠一般的量化财务指标（如运营能力或盈利能力）无法全面评估人工智能初创企业的财务风险。结合人工智能初创企业的特点，本研究还将引入多种定性标准，如投资者地位、技术创新、团队实力、人才激励、市场潜力、竞争环境、人力资源风险等，ER方法在处理定量和定性标准的MCDA问题上独树一帜，因此将在本研究中应用。数据将从2015-2019年期间英国，美国和中国的人工智能初创公司的失败和非失败中收集，训练数据和验证数据将按时间分开。研究对象是提供与人工智能技术高度相关的产品或服务的公司，例如，主要业务收入来自人工智能研发的公司，包括语音识别、计算机视觉、自然语言处理、云计算、传感器、机器人等，将观察超过10，000家公司。此外，研究亦会从每个研究国家（例如中国、美国及英国）选出两间公司作为个案研究，深入访问若干选定公司的经理或专家，以测试模式的准确性。特别是，这项研究将考虑人工智能初创企业是否受到COVID-19的影响，因为企业可能需要数年时间才能从这一大流行病中恢复过来。(1)本研究将通过使用先进的数据分析和ER方法，并结合适当的定性标准，构建一个新的人工智能初创企业财务风险预测模型，以丰富财务风险管理理论。(2)它将提供一个预警机制，帮助AI公司发现财务管理中的问题或缺陷，并提高管理人员的风险意识和能力。此外，它还可以帮助投资者和其他利益相关者理性地投资或与人工智能公司合作。(3)由于人工智能技术对经济发展和社会进步具有重要影响，因此该研究可以支持各国制定人工智能产业规范或技术标准，从而指导人工智能产业健康可持续发展。

AI business has become the most anticipated technology industry and attracts a lot of investment. AI technology is closely related to all aspects of our lives and promotes innovation in traditional industries, including healthcare, education, telecommunications, manufacturing, retail, finance, etc. Moreover, at the national level, it affects citizens' safety and privacy, climate protection, industrial governance and economic policies. Countries are committed to building and developing long-term AI strategies. Therefore, the healthy and sustainable development of AI industry is essential. However, there are many AI start-ups that have become bankrupt due to suffering from financial risks, which has brought adverse impacts on stakeholders and society. Financial risks exist in every part of business management and are affected by various uncontrolled factors, which may result in poor financial status, credit default or even bankruptcy. It is imperative to build a warning model to assess and predict the financial distress of AI firms. Financial risk assessment is in essence a multiple criteria decision analysis (MCDA) problem, aiming to sort many alternatives or select the best solution through information aggregation. This study will use the Evidence Reasoning (ER) approach which is a data-driven machine learning method to predict corporate financial risk. Unlike traditional industries, the operation of AI industry has greater uncertainties, such as larger investment in R & D, more rapid technology update, and higher uncertainty in capital recovery periods, profit models or market demand forecast. General quantitative financial indicators alone (e.g. operational capability or profitability) cannot comprehensively evaluate the financial risks of AI start-ups. Combining the characteristics of AI start-ups, this research will also introduce a variety of qualitative criteria, such as investor status, technological innovation, team strength, talent motivation, market potential, competitive environment, human resource risk, etc. The ER approach is unique in dealing with MCDA problems of both quantitative and qualitative criteria, and will therefore be applied in this research. Data will be collected from failure and non-failure of AI start-ups during 2015-2019 in UK, USA and China, and training data and validating data will be separated by time. The research objects are firms that provide products or services highly correlated with AI technology, for example, firms whose main business income comes from AI research and development, including speech recognition, computer vision, natural language processing, cloud computing, sensors, robots, etc. More than 10,000 firms will be observed. Furthermore, in-depth interviews with a number of selected firms' managers or experts will be conducted by choosing two firms from each of the countries to be studied (e.g. China, USA and UK) as case studies to test the accuracy of the model. In particular, this research will consider whether AI start-ups are affected by COVID-19 since businesses may take years to recover from this pandemic.The contributions of this study are as follows. (1) This research will build a new financial risk prediction model of AI start-ups through the use of advanced data analytics and the ER approach with appropriate qualitative criteria, in order to enrich the theory of financial risk management. (2) It will provide an early warning mechanism to assist AI companies to identify problems or defects in financial management and to improve managers' risk awareness and ability. In addition, it can help investors and other stakeholders to rationally invest in or cooperate with AI firms. (3) Since AI technology has an important impact on economic development and social progress, this study can support countries to make AI industry specifications or technical standards, thereby guiding the healthy and sustainable development of AI industry.