ERL_orderbook (1) - Evolutionary Reinforcement Learning in...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Evolutionary Reinforcement Learning in FX Order Book and Order Flow Analysis R G Bates, M A H Dempster and Y S Romahi Centre for Financial Research, Judge Institute of Management, University of Cambridge Emails: rgb2@cam.ac.uk, mahd2@cus.cam.ac.uk, yr206@cam.ac.uk Abstract: As macroeconomic fundamentals based modelling of FX timeseries have been shown not to fit the empirical evidence at horizons of less than one year, interest has moved towards microstructure-based approaches. Order flow data has recently been receiving an increasing amount of attention in equity market analyses and thus increasingly in foreign exchange as well. In this paper, order flow data is coupled with order book derived indicators and we explore whether pattern recognition techniques derived from computational learning can be applied to successfully infer trading strategies on the underlying timeseries. Due to the limited amount of data available the results are preliminary. However, the approach demonstrates promise and it is shown that using order flow and order book data is usually superior to trading on technical signals alone. Keywords: Reinforcement learning, evolutionary learning, FX time series analysis, order flow analysis 1. INTRODUCTION Published work directly relevant to the study of orders and transaction flows has appeared in relation to equity markets, but much less so for foreign exchange markets. Clearly the previous lack of data on orders and transactions in the FX market is responsible for the lack of such work, while the greater availability of data in the equity markets has produced more work of relevance (see [3]). Early work on foreign exchange markets employed a macroeconomic approach, attempting to explain movements in exchange rates in terms of macro variables such as balance of payments, interest rate differentials, inflation differences (Purchasing Power Parity), etc. Even on timescales of a month these methods have little predictive power [16,17]. To quote Frankel & Rose [10]: “the Meese and Rogoff analysis at short horizons has never been convincingly overturned or explained. It continues to exert a pessimistic effect on the field of empirical exchange rate modeling in particular and international finance in general”. Subsequent work [3,15] demonstrated that macro models begin to explain some exchange rate variation at horizons over 1 year. In foreign exchange markets a number of studies have looked at Central Bank intervention and its effects. Dominguez [7] looks at G-3 central bank intervention, treating it as an information source. Using high frequency data the effectiveness of the intervention on the exchange rate and its volatility is assessed. The study shows that intervention is most effective when timed near major macro announcements and during periods of heavy trading volume (note not, as might be expected, during periods of low liquidity). Post-intervention mean reversion in both exchange rate and its volatility is observed.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 12/08/2011 for the course CIS 625 taught by Professor Michaelkearns during the Spring '12 term at Pennsylvania State University, University Park.

Page1 / 8

ERL_orderbook (1) - Evolutionary Reinforcement Learning in...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online