Author: Sriram Kashyap
Degree: M.S. in Financial Engineering
Year: 2020
Advisory Committee: Dr. Ionut Florescu, Dr. Dragos Bozdog
Abstract: High frequency trading algorithms are purpose built to produce Buy/Sell signals at specific volume based on predefined rules. These signals are then converted to market or limit orders and routed to exchanges. The rules to come up with the Buy / Sell signal are often static and are designed before hand in a Trading strategy. During the trading day however, there might be time periods when the employed strategy is not the most profitable one. This thesis proposes an approach to introduce a new kind of signal- Alter strategy signal to modify the strategy which is generated by continuous learning from the measures of the limit order book during the trading. Ability to learn the market condition and switch between strategies can be beneficial to:
- Profit from regime switches in the Limit Order Book
- Recover from Impact of Trading activity
- Explore alternative strategy opportunities
By measuring the Limit Order Book characteristics in the learning period and
comparing with the current time characteristics we increase the situation awareness of the strategy as a whole and provide more options to respond to market events, as compared to - Hold, Buy or Sell options that the strategies have classically had.