Haitong International makes waves in artificial intelligence

Award-winning Hong Kong-based Chinese securities house embraces AI to develop proprietary trading models, build better predictive systems and go that extra mile to deploy cutting edge technology for international clients.

Haitong International makes waves in artificial intelligence

In the 1990s business leaders across all industries tried to work out what the rise of the Internet would mean for their businesses. Move forward 20 years, and the same thing is now happening with artificial intelligence (AI).

Given its huge potential to offer competitive advantage, existing AI success stories have not gone unnoticed. About nine financial service companies out of 10 are now working with AI-based technology, according to Deloitte and industry group EFMA data. Research indicates that US$11 billion has been invested into AI since 2010 and that figure is expected to rise to more than US$47 billion by 2020.

AI TRADING AT HAITONG INTERNATIONAL

To better execute clients’ orders, Haitong International Securities Group (665.HK) is ramping up its algorithmic execution services. In June 2017, it became the first Hong Kong-based Chinese financial services provider with proprietary electronic trading platform designed for its global institutional clients.

The newly expanded offering comprises sales trading, portfolio trading, algorithmic trading and direct market access (DMA). Haitong International currently covers major global financial centres including Hong Kong, US, Japan and Australia, and is expected to extend to major stock exchanges in Europe and Asia-Pacific in 2018.

While traditional algorithmic execution services excel at speed and scalability, Haitong International’s next-generation platform, which is capable of predicting the movement of stock price and trading volume by taking advantage of the latest AI techniques such as neural network and deep learning, raises the industry standards. AI and quantitative analysis play such an important role in Haitong International’s execution services that its algorithmic trading team has one of the highest ratios of quant and analytics personnel, relative to traders and salespeople, among its Asia-based peers. In 2017, Haitong International’s trading team has grown to 30 from 11 with latest hires from CIMB, Jefferies Group, and UBS.

1. Haitong International developed proprietary prediction models in its algorithmic trading service

Within the electronic execution space, the largest single determinant of performance is the quality of the predictions that the strategy has access to. Inevitably the price forecasts have the biggest impact, but predictions of other aspects of trading such as traded volume, volatility and the volume curve are also very significant.

The old non-AI approach forecasting these features was simply to use averages (or average profiles) over multiple trading days. For key features, such as closing auction volume, the averages can be improved upon using carefully hand-crafted models. The creation and evaluation of these models is a slow, time consuming process, often taking many months of a quantitative analyst’s time.

2. AI enables Haitong International to build better prediction models

Deep Learning, a class of machine learning algorithms that uses multiple layers of processing units, enables Haitong International to significantly speed up the process, achieving similar
(and sometime much improved) results but at much faster speeds. The approach allows the company to develop models for all features of interest to trading, moving away as much as possible from static averages that do not adapt to the observed trading on the day. It also enables the company to run completely retrained models for different markets, with the prediction potentially relying on completely different features – something that cannot be attained with hand-crafted models. Finally, these models come with the significant bonus that the models are completely dynamic, being constantly updated with the latest trading data and being able to adapt to structural changes in the market without any human intervention.

3. Haitong International is applying the state-of-art AI techniques to go even further

At the current stage, the company is reaping significant performance benefits from simply improving the quality of the input data to existing execution strategies, but machine-learning techniques give it a path to go much further. Reinforcement learning and dynamic programming techniques allow training of complicated strategies that have a huge range of possible actions. This is of particular interest to execution trading, where the determination if a particular action was good or not is only possible at the end of trading. These approaches are being evaluated internally at the moment, and Haitong International anticipates that they will drive the next generation of execution strategies.

HAITONG INTERNATIONAL'S VISION FOR AI

In addition to algorithmic trading, Haitong International believes that AI has an important role to play in many aspects of financial services. The company is adopting AI-enabled processes and systems for a wide variety of areas. To facilitate this transformation, Haitong International established last November the Global Markets Artificial Intelligence Trading Committee to oversee the firm-wide effort of deploying AI technologies to help its staff become better bankers, traders, sales, and executives. Haitong International is committed to becoming a leading technology financial services institution by continuing to invest in quantitative trading, algorithmic trading and smart trading technology platforms.

Haitong International believes firmly that the future of investment banking business lies in relationships and in the valuable services that financial services institutions can bring to clients, whether individuals or corporates. The company believes that AI will allow it to free its people from routine work and allow them to concentrate on more creative, value-added services. Haitong International thinks that AI will increasingly become a competitive advantage for itself in unlocking new markets and increasing efficiency.

Numerous AI applications can now be seen in modern investment banking businesses: What Haitong International is more focused on now is AI’s application within sales and trading. Today, nearly 45% of the revenue generated from cash equities trading comes from electronic trades, according to Coalition, a UK research firm that tracks the industry. On top of fruitful success achieved by its electronic trading team, Haitong International will be ready for reaping the opportunities that come along with the arrival of the so-called “intellectual investment banking era” with its joint effort across business lines and innovative transformation in traditional business models.

In order to help investors ride on the AI waves, Haitong International will hold an AI conference in the second half of 2018 with industry leaders, government officials, entrepreneurs, investors and academics - from across the region and around the world to share their forward-thinking insight.

Print Edition

FinanceAsia Print Edition

EVENTS