AI model analyzes mountains of financial data to help market traders make better buy-and-sell calls.

The goal—and key challenge—of every fund manager and trader is to achieve higher returns for their investor clients.

While many used quantitative trading methods and already analyzed data like price and volume, financial data is notoriously “noisy,” meaning there are many potentially misleading signals that need to be filtered out.

The sheer volume of data and the fact that it can all fluctuate in real-time also means that even the most meticulous traders can find it hard to keep up or may only use a fraction of all the data that is out there.

Fund managers and traders could soon be making better informed buy-and-sell market decisions and bring in higher returns for their investor clients with the help of an artificial intelligence (AI) model designed to sift through and analyze vast amounts of real-time financial data.

China Asset Management Co., Ltd. (China AMC) and Microsoft researchers have developed the model that is now undergoing testing.

Trials of the AI model started under real market conditions in China in March. The results have been encouraging to date and well ahead in performance compared with the progress of the market or that of specific indexes.

The initial success has raised hopes within the company that AI could potentially revolutionize decision-making and boost investment performance without changing its existing business and operational modes.

“AI is a critical technology in driving financial transformation, and it is of great significance to tap into the convergence of AI and financial services,” says Li Yimei, China AMC’s general manager.

China AMC got together with Microsoft Research Asia (MSRA) through the latter’s “Innovation Partnership” program. MSRA used the program to share its AI knowledge and expertise with industries and sectors to help them develop new solutions and drive digital transformation.

In this case, China AMC wanted to introduce AI to the discipline of Quantitative Investment, which aimed to deliver higher returns.

A new way to handle a trusted method

Quantitative investment isn’t new. Put simply, it is a method of analyzing data – like price and volume. For instance, it calculates which stocks to buy or sell and when, using a systemic approach that should be immune from the emotional ups and downs of market sentiment.

The tricky part is collecting the right data from the mountains of information available. Right now, even meticulous quantitative traders may only use a fraction of all the data that is out there.

Additionally, financial data is notoriously “noisy,” meaning there are many potentially misleading signals that need to be filtered out.

This is where AI comes in. It’s good at dealing with vast amounts of dynamic data in real-time and applying complex logic. Using machine learning, China AMC and MSRA developed AI algorithms that use two sophisticated methods to predict the market.

The first is called a “spatio-temporal convolutional neural network.” It identifies patterns in volume-and-price data by estimating what is going on among the different variables and isolating important factors to watch.

The second is called a “time-varying attention model.” It compares those factors with what is going on in the market at any given moment. It then combines them in a time-varying way to identify dynamic market trend patterns. This results in an α signal – which is a measure of how well an asset is performing.

Empowering financial professionals to make better decisions, not to replace them 

Importantly, the system’s developers avoided using a learned model approach, which could reinforce data bias – a problem that has been a typical challenge for other AI systems.

Tie-Yan Liu, assistant managing director of MSRA explains: “The financial market is highly dynamic – a fixed investment formula faces great challenges when dynamic data is changing all the time. Our new AI model can identify several factors in real-time change, and it can extract α signals in a timely way to update the model. This ensures we have an adaptive investment formula that is always at its best at any time.”

The result is a more profitable investment strategy called “AI+ Index Enhancement.” In effect, it is a tool that seasoned traders can use. Combined with their experience and specific domain knowledge, it empowers China AMC’s financial professionals to make better decisions, not to replace them.

AI-based intelligent investment can help fund managers and traders overcome bias that can be created by people’s personalities or emotions.

It can also provide a stable reference for financial professionals to judge the market better.

Decisions are made by people with experience

But in the end, decisions are made by people who also relied on their own industry experience. AI is meant to augment the capabilities of professionals to provide more possibilities and services in the financial sector.

As Zhang Hongtao, managing director of AMC and head of China AMC quantitative investment division said: “AI models fit well into our investment strategies at China AMC. And, they are also working smoothly with other ‘components’ at China AMC. Without a single change to the business and operation modes, AI is already able to revolutionize the area of intelligent investment.”