Feature Engineering for Mid-Price Prediction With Deep Learning

Mid-price movement prediction based on the limit order book data is a challenging task due to the complexity and dynamics of the limit order book. So far, there have been very limited attempts for extracting relevant features based on the limit order book data. In this paper, we address this problem by designing a new set of handcrafted features and performing an extensive experimental evaluation on both liquid and illiquid stocks. More specifically, we present an extensive set of econometric features that capture the statistical properties of the underlying securities for the task of mid-price prediction. The experimental evaluation consists of a head-to-head comparison with other handcrafted features from the literature and with features extracted from a long short-term memory autoencoder by means of a fully automated process. Moreover, we develop a new experimental protocol for online learning that treats the task above as a multi-objective optimization problem and predicts: 1) the direction of the next price movement and; 2) the number of order book events that occur until the change takes place. In order to predict the mid-price movement, features are fed into nine different deep learning models based on multi-layer perceptrons, convolutional neural networks, and long short-term memory neural networks. The performance of the proposed method is then evaluated on liquid and illiquid stocks (i.e., TotalView-ITCH US and Nordic stocks). For some stocks, results suggest that the correct choice of a feature set and a model can lead to the successful prediction of how long it takes to have a stock price movement.