Intern/Working experiences

Investment research intern at Acadian Asset Management

Project 1

In this project, I managed to propose a risk-free strategy by find exact-match of stocks from different brokers. The annualized risk-free rate of the strategy can be 2% with notional value of 10 millions. I also proposed a strategy of trading specific stocks (with high compension from brokers) and managed to reduce systematic risks by solving quadratic optimization problem.

Project 2

In the second project, I first analyzed the relationship between companies’ earning surprise with extra returns. And then predicted earnings’ surprise from a classification perspective. I porposed signal processing methods to pull information from time series before earnings announcement. I also designed comparison framework in choose important features that can be applied into prediction model.

When doing prediction, I tried different tree-based methods (e.g. XGBoost, Random Forest, etc) to avoid the “black-box” of deep learning method and achieved a testing AUC of over 0.8. Meanwhile, I also compared tree-based methods with convolutional neural network method to assess the performance of my model. Finally, I also proposed a new framework of convolutional neural network with specific structure that is more transperant and explanable. The model framework is well-defined and can be transfered into solving other prediction problem.