Research

Research Projects

• Macroscopic look at the Equity Markets

This research aims to investigate the existence of sensors which may aid in monitoring equity markets. While the classical approach consists of following the evolution of individual stocks, we examine equity markets macroscopically. This method enables us to extract significant information about the overall market dynamics. This way, we see the market within the context of the principles of mass and momentum conservation and the variables density, flux, pressure, average velocity, etc. We can then define "sensors" that can monitor some of these variables in the market. In the era of high-frequency trading, a powerful tool is needed to monitor market activities. In this research, we propose a method to predict unusual activities and provide alerts. Our analysis shows that our sensors are able to predict and capture valuable information about the flash crash day, when the volatilely was extraordinary high.

• Prediction and Uncertainty Quantification of Wind Power

We develop an integrative framework to predict the wind power output, considering many uncertainties. For probabilistic wind power forecasts, all the sources of uncertainties arising from both wind speed prediction and wind-to-power conversion process should be collectively addressed. We model the wind speed using the inhomogeneous geometric Brownian motion and convert the wind speed's prediction density into the wind power density in a closed-form. The resulting wind power density allows us to quantify prediction uncertainties through prediction intervals and to forecast the power that can minimize the expected prediction cost with unequal penalties on the overestimation and underestimation. We evaluate the predictive power of the proposed approach using data from commercial wind farms located in different sites. The results suggest that our approach outperforms alternative approaches in terms of multiple performance measures.

• An Alternative Prediction Approach Based on Real Option Theories

This research introduces a new prediction model for time series data by integrating a time-varying Geometric Brownian Motion model with a pricing mechanism used in financial engineering. The proposed model is useful when overestimation needs to be penalized differently from underestimation. We apply the pricing mechanism to quantify under and over-estimation costs and by employing a weight parameter in the prediciton cost function, we adjust the forecast results based on the prediction preference. We evaluate the approach using three real datasets. The numerical results demonstrate that the proposed model shows competitive prediction capability, compared with alternative approaches including Auto-Regressive Integrated Moving Average (ARIMA) and General Auto Regressive Conditional Heteroskedasticity (ARIMA-GARCH) models.

• Deep Learning in Finance

The limited success of econometric models in making financial predictions especially during highly volatile periods, such as when reacting to the announcement of Brexit, has created a need to look at more novel techniques. With the recent advent in technology, learning from highly complicated data has become more feasible. In this research, we explore the use of machine learning algorithms such as Deep Neural Networks (DNNs) and Support Vector Machines (SVMs) in testing a hypothesis about the behavior of equity markets before and after the Brexit announcement. We use minute by minute data to test if a systematic pattern is observed in the flow of information from high-value to low-value stocks, aiding financial prediction.