Dr. Ionut Florescu
Director, The Hanlon Financial Systems Lab
Ionut Florescu, Ph.D., is Research Associate Professor in Financial Engineering at Stevens Institute of Technology and Director of the Hanlon Financial Systems Laboratories.
Professor Florescu expertise lies in developing stochastic models and using them for real-life applications. As detailed in many in publications, these applications pertain to computer vision, cryptography, environmental studies, geophysics and transformative learning.
One such application is the Stevens High Frequency Trading (SHIFT) Simulation System which started development in January 2014. It is the first model of its kind to test the behavior of modern high frequency financial markets using live, real-time market data. Other applications include the ABCShift a patented computer vision algorithm that allows tracking of objects in videos when the background is changing, RAPID - the cloud robotics platform used for developing application, liquidity studies in finance and stochastic volatility modeling.
Doctor of Philosophy in Statistics, Purdue University, West Lafayette, Indiana, USA
December 2004. Primary Research Area: Mathematics of Finance
Master of Science in Statistics, Purdue University, West Lafayette, Indiana, USA
December 2001. Specialization in Computational Finance
Master of Science in Mathematics, University of Bucharest, Romania
July 1997. Specialization in Stochastic Processes
Stevens Institute of Technology, Department of Mathematical Sciences, U.S.A.
Research Associate Professor and Director of the Hanlon Financial Systems Lab 2012 - present Assistant Professor Fall 2005 - Spring 2012
Purdue University, Department of Statistics, U.S.A.
Visiting Assistant Professor, Spring 2005 Teaching Assistant, Fall 1998 - Fall 2004
Romanian Academy, Center for Mathematical Statistics, Bucharest, Romania
Research Assistant, Fall 1997-Spring 1998
University of Bucharest, Department of Physics, Romania
Lecturer, Fall 1997 - Spring 1998
Professor Florescu grew up in Romania during its Communist-ruled years. In high school mathematics became his passion and he participated in mathematical Olympiads. A graduate of Mathematics Division at University of Bucharest he went on to earn a Ph.D. in statistics from Purdue University. He believes fervently that teaching is his calling. His is a family of teachers, his grandfather was a famous geology professor who contributed to development of the Romanian Atlas series. Coming from a very wealthy and influential family before 1945 his grandfather endured the prosecution of a political system that assigned him to teach in an elementary school and ironically a new school for the ruling class (Academia Stefan Gheorghiu). Today, Professor Florescu remembers days on the streets of Bucharest when people would recognize his grandfather—their former geology teacher—and recount him with stories of what a great teacher he was. Professor Florescu resides with his wife and boy in Hoboken, NJ.
Professor Florescu has spent all his academic career at Stevens Institute. Starting in 2005 as an Assistant Professor in the Department of Mathematical Sciences he then moved to the Financial Engineering Division in 2012. He developed and introduced multiple courses in the FE division as well as two certificates and a new Master degree. He is the recipient of several grants from NSF, Nvidia, and CME foundation. He organized several conferences, including the series of conferences on Modeling High Frequency Data in Finance at Stevens Institute (PUT LINK). He serves as a reviewer for over 30 journals.
Master's Thesis Students
He is an expert in developing stochastic models and using them for real life applications.
In this project we are analyzing the particular details specific to markets in mainland China and Hong Kong. Due to the particular regulations existing in these markets and the inherent introduction of the option exchange it created very interesting situations in these market where new models need to be created.
Volatility is the market is a process that clusters and has long memory. Both these features may be accomplished by modeling volatility using a fractional Brownian motion. However, estimating the Hurst parameter which completely characterize this process using financial data is a problem that needs to be solved before these models are implemented in practical applications. This project uses a filtering technique we are developing for this purpose.
This is a project where we study implementations and parallelization of our algorithms. Applications we have already implemented are the quadrinomial recombining tree used to price any stochastic volatility model, and estimation of integrated fractional Brownian motion.
In this project we have created a methodology to detect times during the trading day where we have unusually large price movement with very little associated volume. During the course of the project we have observed that the detected events tend to associate with a lack of liquidity in the market. Thus the project is now documenting the connection and at the same time is attempting to create a liquidity index for each asset on the market.
This is a project that is attempting to model a real exchange to provide realistic answers to questions of current interest. For example, we will be able to answer the question of which regulation will be detrimental to markets and which will be beneficial, study of impact of high frequency trading during various market conditions, and much more.
This project is related to the previous one but it is focusing on the visualization of the resulting estimated volatility process. The data is multidimensional and to better serve as inputs it needs to display the relevant information in a manner that is easy to understand and point out opportunities.
In this project we concentrate on developing stochastic volatility models for better pricing of these OTC instruments. We are drawing on our expertise with these models and our previously created methodology of approximating them using a parallelizable algorithm. These are long term instruments and a very dynamic volatility process is needed to cope with the possible risk during such long intervals. The current models used (geometric Brownian motion, local volatility, etc.) are not appropriate for these instruments and the complex models do not have an analytical solution.