In January 2020, the center for alternative Finance (CCAF) of Cambridge University released the results of a study on the impact of artificial intelligence on the financial industry. The study is one of the most comprehensive global surveys in the industry. The respondents include 151 institutions from 33 countries, including existing financial institutions and financial technology companies.
The results are as follows: at least 77% of respondents believe that AI will have a significant impact on their company in the next few years. About 64% intend to use AI for new customer development, customer service, risk management, process automation and new product development. At present, AI has been widely used in risk management, with a utilization rate of 56% in the company.
In the financial field, high-frequency trading companies and hedge funds have always been the leaders in the application of AI. Recently, financial technology companies, insurance companies, banks and regulators are catching up. AI applications in the industry include robot consultant, backtracking test, model verification, portfolio composition and optimization, stress test, algorithm trading and law compliance. Next, let’s introduce these applications in detail.
AI and machine learning algorithms are gradually subverting financial risk management. AI driven solutions provide the following insights:
1,Determine the loan amount of the customer.
2,Alerts traders about risk positions.
3,Enhance compliance and limit model risk.
For a long time, investment companies have used computers to complete transactions. A large number of hedge funds rely on data scientists to build data models. However, this method has a major limitation – it only uses historical data that are mostly static and rely on human intervention. Therefore, these calculations are difficult to change according to market changes.
Modern AI models use algorithmic trading to achieve big strides. The difference between these models is that they not only analyze massive data, but also the analysis process is completely automated – the models are continuously studied and improved, and finally can be comparable to human beings. This “intelligence” comes from complex machine learning technologies, such as evolutionary computing (genetic search algorithm) and Bayesian networks.
AI tools draw massive data from resources all over the world, learn from them and give corresponding predictions. The data is collected thoroughly: it extracts information from financial exchanges, news reports, books, social media platforms (such as twitter) and even TV programs.
It is important to understand how AI goes deep into the financial field. It is different from the traditional technical intervention that allows humans to determine financial strategies. Here, AI is the dominant.
Another application of AI in the financial field is the rapid development of fraud detection. The cybercrime industry stole about $600 billion, accounting for about 0.8% of the total GDP created by global commerce. Cyber criminals are becoming more and more shrewd and use modern technology to achieve their criminal purposes. Statista predicts that the growth of the integrated anti fraud and prevention market will exceed $40 billion by 2022.
So how does AI work? Through AI, modern machine learning, cyber criminals integrate supervised and independent technology to establish a model with accurate prediction ability.
Supervised learning uses annotated data (data that is manually processed and considered fraudulent activity) and learns complex patterns from corporate data sets. At the same time, unsupervised learning processes previously unrecognized data and infers the data structure by itself. Other anti fraud methods include regression and classification. Both can analyze data to determine whether a transaction is fraudulent.
Standard supervisory algorithms for solving problems include:
1,The decision tree helps to introduce a set of rules that train fraud instances and learn normal customer behavior, so they can identify abnormal situations and give alerts.
2,Based on human brain, neural network can learn and adapt to customer behavior to monitor real-time fraud.
Unsupervised learning algorithms include:
1,K-means clustering divides the data set into similar data point sets (clusters) for anomaly detection.
2,The local outlier factor determines the local density of data points and identifies areas with similar density. Data scientists can use the concept of locality to mark points with abnormally low density, called outliers. This method can be used to detect fraudulent transactions.
Compliance is an important function of finance, especially in the current economic crisis. Follow the functions related to enterprise risk management and dealing with risks, such as operational, market and credit risks.
Regulatory technology is an advanced function related to compliance in the field of financial technology. Here, when it is used to continuously monitor the activities of a company, AI can reflect its advantages. In this way, it provides valuable real-time insight to prevent the recurrence of violations at the first time. In addition, this regulatory model allows companies to save regulatory capital and use automation to reduce excessive compliance costs – large financial companies spend up to $70 billion a year on compliance.
Reading and interpretation of legal documents (especially for determining meaning). Waymark, based in London, provides this service to financial companies. Although AI has many applications in the financial field, it also has disadvantages. The industry needs to correct practical problems to enhance the application of AI.
One of the biggest concerns is the availability of appropriate data. Although R language and python can read any form of data from excel tables to SQL / NoSQL datasets, AI driven solutions still run slower than enterprises’ ability to accurately organize their internal data. Usually, data is stored in different silos of different systems in various departments, and the regulatory and internal political difficulties of different systems limit information sharing.
Another difficulty is the lack of skilled talents: not only master expert AI, machine learning and data science knowledge, but also have experience in building and applying AI centered solutions in the financial industry. Ray Kurzweil once predicted: “artificial intelligence will reach the level of human intelligence around 2029. Further, by 2045, we will expand the capacity of intelligent technology, that is, biological machine intelligence created by human civilization, by 1 billion times. ”
With the cooperation of human and AI, there will be infinite imagination in the future.
Leave a message to praise and pay attention. Let’s share the dry goods of AI learning and development.