Our program actively invites academic and industry visitors to supervise students in cutting-edge research. We have recently launched capstone project courses to collaborate with financial firms (including Morgan Stanley, JP Morgan, Founder Securities, WorldQuant and others).  The first-hand insights from supervisor will benefit students greatly and potential employers could take a first look at the talent in the process.

Project highlights

Towards a Characterization of the Variability of Trade Arrival Processes

In algorithmic trading, researchers have often described trading as clustered or 'bursty' yet few have defined burstiness. In this research we propose to measure how trades are conducted in three ways, namely Poisson, Batch-modulated Poisson, and Markov-Modulated Poisson. An understanding of how and when to trade is critical to uncover of the price formation process, and will have many important applications from risk management to algorithmic trading. We will also examine how the seasonality of the trading day might be incorporated into this models.

The Hurst Exponent in Finance

The degree to which long memory is involved in the formation of price return, particularly pertaining to Asian equities, is important, for it may allow us to trade stocks with long memory differently from those without. We propose to quantify, for the Nikkei, a measure of the long memory of the time series in question: specifically, the Hurst Exponent. We will also investigate the degree to which long memory is related to other microstructural factors, primarily bid-offer spread.

Modelling the Impact of Trades in Asian Markets Using a Power-Law Model

Power-law models are important in high frequency models, in particular when it comes to the modeling of trade data. We propose, using Asian trade data, to analyze the impact of trading using these models by taking into account the scale of imbalance on the orderbook. We will test the applicability of our results by incorporating the model in a trading simulation.

Chinese Stock Entity Recognition System
Some financial search engines can find the latest financial news and relevent information for a given stock effectively and efficiently. This project focus on identifying the relevant stock entities given a piece of news. The technology we employ is deep learning based natural language processing method. The ideal delivered outcome is a system which can automatically extract the stock entities for the real-time live Chinese news.
Robo-Advisory (Model Portfolio Construction)
Model Portfolio is a key attribute of Robo-advisory. A huge amount of data crunching is required between fund categorization, theme definition and goal matching before the actual portfolio optimization could be carried-out. This project is to conduct a series of model portfolios to deal with clients’ demand or market’s investment trend with a better diversified, risk-reward approach. Advance capabilities on model portfolio is to be constructed by introducing model portfolio in multi-dimensions and multi-life goals.
RFP Style Analysis and Company Style Box
To construct a multi-dimensional style box to reflect a fund’s attributes on both risk and return in different market cycles, this project redesigns the RFP to include more statistical friendly items by facilitating the analysis on the management style and aptitude. We analyze fund manager based on RFP which is an investment questionnaire collecting information on management style. The questionnaire includes the manager’s idea sourcing decision making process, selection quiteria (Stock/Bond), sell discipline/cut-loss mechanism…etc. In this project, Company Style Box is developed to allow multi-dimensional analysis of funds’ performance style.