Title: Analysis of high-frequency order flow dynamics with machine learning techniques
Authors: Martin Magris - Tampere University of Technology (Finland)
Adamantios Ntakaris - Tampere University of Technology (Finland) [presenting]
Juho Kanniainen - Tampere University of Technology (Finland)
Alexandros Iosifidis - Tampere university of technology (Finland)
Moncef Gabbouj - Tampere university of technology (Finland)
Abstract: We use machine learning techniques to train a predictive model based on ultra-high frequency limit order flow data to capture and predict the short term dynamics of the order book. Our ITCH feed data allows to precisely calibrate multi-class learning models to forecast several quantities such as the mid-price movement and spread crossing. Both training and prediction of our classifiers rely on a number of features designed to characterize the order book. In our application, we address the issues of effective features selection, transformation and model calibration. Based on results for several equities traded in Nasdaq Nordic market, we discuss the effectiveness of this approach in predicting books dynamics.