Forecasting limit order book price changes using change point detection

Abstract: The main purpose of this thesis is to propose a method for using a change point detection algorithm to forecast short term limit order book price changes. The idea is to test whether a significant change of the shape of the limit order book contains any information about impending changes to mid market. Using a data set consisting of all the changes of the limit order book throughout the trading day, a change point detection algorithm is used to detect what is deemed to be significant changes of the shape of the limit order book. A measurement of the limit order imbalance is constructed as a proxy for the shape of the order book which then is used as input signal to the change point detection algorithm. A new data set is created based on the detected change points and a regression is run based on these to forecast price changes. It is found that the change point data set contains a certain amount of information about impending price changes.