This type of machine learning is referred to as supervised learning, where algorithms use historical data to learn from and identify patterns that may be of interest to someone investigating fraud. In reality, fraud detection uses a combination of data science methods, from supervised to unsupervised learning - where the goal is to help find and identify previously unknown patterns in the data. Modern fraud detection practices will almost always boast about the use of some form of machine learning. For example, the top three fraud detection software programs - "Signifyd", "Riskified" and "Kount" - all mention the use of various machine learning methods according to G2, the world's largest tech marketplace [7].
这种类型的机器学习被称为监督学习，其中算法使用历史数据来学习和识别欺诈调查人员可能感兴趣的模式。实际上，欺诈检测使用数据科学方法的组合，从有监督学习到无监督学习，其目标是帮助发现和识别数据中以前未知的模式。现代欺诈检测实践几乎总是吹嘘使用某种形式的机器学习。例如，根据全球最大的技术市场G2，排名前三位的欺诈检测软件程序——“Signifyd”、“Riskified”和“Kount”都提到了各种机器学习方法的使用[7]。
Forecasting and Predictive Analysis 预测与预测分析
Another key application of data science in banking and finance is forecasting when and how markets will move - and predictive analysis is used to do this. Predictive analysis is the use of data, statistical algorithms and machine learning techniques to identify the likelihood of future outcomes based on historical data [8]. This is not something new to the industry - for example, life insurers have been using mortality statistics for decades in order to make predictions for when policyholders may die. However, with the continual increase in the volume and types of data and computers becoming faster and cheaper, this technology is about to enter its golden age.
数据科学在银行业和金融业中的另一个关键应用是预测市场何时以及如何变化——预测分析就是用来实现这一点的。预测分析是利用数据、统计算法和机器学习技术，根据历史数据确定未来结果的可能性[8]。这对保险业来说并不是什么新鲜事——例如，寿险公司几十年来一直在使用死亡率统计数据来预测投保人的死亡时间。然而，随着数据量和类型的不断增加，以及计算机变得更快、更便宜，这项技术即将进入黄金时代。
Many investment banks and financial institutions now employ predictive analysis to model financial markets. One method is to use genetic algorithms - a problem-solving method that mimics the theory of evolution [9] - to find the best combination of parameters and biases in a given trading rule. These algorithms repeatedly test a variety of parameter value sets, each with slight differences. At each step, the algorithm will randomly select a number of parameter value variants and allow these to produce 'children', which over successive generations will 'evolve' the parameter set to an optimal solution. This means that traders, whether that be large institutions or individuals, are able to create models based on initial parameter sets and optimise them using genetic algorithms and historical market data. By making advanced trading and analytical knowledge accessible to people other than the experienced industry experts, there is the potential to disrupt not only the landscape in terms of who trades on the markets but also to the actual financial market itself. Every entity from the large multinational quantitative trading company to the sole trader in their living room has the opportunity to create successful algorithms to help them trade and make a profit, which has the potential to massively change the accepted trends in how the markets rise and fall over time.
许多投资银行和金融机构现在采用预测分析来模拟金融市场。一种方法是使用遗传算法——一种模仿进化理论的问题解决方法[9]——在给定的交易规则中找到参数和偏差的最佳组合。这些算法反复测试各种参数值集，每个参数值集都略有差异。在每个步骤中，算法将随机选择多个参数值变量，并允许这些变量产生“子代”，这些子代将在连续几代中“进化”参数集以获得最佳解决方案。这意味着交易员，无论是大型机构还是个人，都能够基于初始参数集创建模型，并使用遗传算法和历史市场数据对其进行优化。通过向经验丰富的行业专家以外的人提供先进的交易和分析知识，不仅有可能破坏市场上交易人员的格局，也有可能破坏实际金融市场本身。从大型跨国定量贸易公司到起居室中的独家贸易商，每个实体都有机会创建成功的算法来帮助他们进行交易并获利，这有可能极大地改变市场随时间起伏的公认趋势。
Conclusion 结论
We have seen some examples of areas in finance and banking that have changed dramatically thanks to the introduction of data science. The quantitative nature of finance and the vast amounts of data that are involved when considering financial markets, global transactions, personal banking and investment management make this industry a prime target for innovation through data. Many companie
