Sequence classification of the limit order book using recurrent neural networks

Dixon, Matthew

doi:10.1016/j.jocs.2017.08.018

Cited by 65 publications

(28 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The author in [48] develops a new type of deep neural network that captures the local behavior of a LOB for spatial distribution modeling. Dixon applies RNN [20] on S&P500 E-mini futures data for a metric prediction like price change forecasting. Minh et al [38] also propose RNN architecture for short-term stock predictions by utilizing successfully financial news and sentiment dictionary.…”

Section: Literature Reviewmentioning

confidence: 99%

Feature Engineering for Mid-Price Prediction With Deep Learning

et al. 2019

View full text Add to dashboard Cite

Mid-price movement prediction based on 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 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 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 i) the direction of the next price movement and ii) 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.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Feature Engineering for Mid-Price Prediction With Deep Learning

et al. 2019

View full text Add to dashboard Cite

show abstract

“…We would like to point out that the main idea of the present work is to utilize the majority of the technical indicators 2 and provide a fair evaluation against other state-of-the-art features and advanced quantitative hand-crafted features like the adaptive logistic regression feature and see which ones are more informative. Classification performance can be easily improved by utilizing more advanced classifiers like convolutional neural networks and recurrent neural networks (e.g., [22]), but it is outside Table 9: List for the first 10 best features for the 5 sorting methods of the scope of our evaluation. Our work open avenues for other application as well.…”

Section: Sortingmentioning

confidence: 99%

Mid-Price Prediction Based on Machine Learning Methods with Technical and Quantitative Indicators

et al. 2018

View full text Add to dashboard Cite

Stock price prediction is a challenging task, but machine learning methods have recently been used successfully for this purpose. In this paper, we extract over 270 hand-crafted features (factors) inspired by technical and quantitative analysis and tested their validity on short-term mid-price movement prediction. We focus on a wrapper feature selection method using entropy, least-mean squares, and linear discriminant analysis. We also build a new quantitative feature based on adaptive logistic regression for online learning, which is constantly selected first among the majority of the proposed feature selection methods. This study examines the best combination of features using high frequency limit order book data from Nasdaq Nordic. Our results suggest that sorting methods and classifiers can be used in such a way that one can reach the best performance with a combination of only very few advanced hand-crafted features.

show abstract

“…Following Kercheval and Zhang (2015), Sirignano (2016), Dixon (2017), we compose our feature set of five levels of prices, volumes, and number of limit orders on both the ask and bid side of the book. We additionally, and somewhat heuristically via a process of "feature engineering," characterize order flow by the ratio of the number of market buy orders arriving in the prior 50 observations to the resting number of ask limit orders at the top of book.…”

Section: High-frequency Datamentioning

confidence: 99%

“…It is designed to provide predictors in complex settings where relations between input and output variables are nonlinear and input space is often high dimensional. A number of researchers have applied machine learning methods to the study of limit order book dynamics (Dixon, 2017;Dixon, Polson, & Sokolov, 2017;Kearns & Nevmyvaka, 2013;Kercheval & Zhang, 2015;Sirignano, 2016). DIXON…”

Section: Introductionmentioning

confidence: 99%

“…This article takes an algorithmic approach to predicting the next event price-flip from a short sequence of observations of limit order book depths and market orders. We choose a spatiotemporal representation (Dixon, 2017;Sirignano, 2016) of the limit order book combined with history of the market orders as the predictors. Our approach solves a sequence classification problem-a short sequence of observations of book depths and market orders can be classified into directional mid-price movement.…”

mentioning

confidence: 99%

See 1 more Smart Citation

A high‐frequency trade execution model for supervised learning

Dixon

2018

High Frequency

Self Cite

View full text Add to dashboard Cite

This article introduces a high-frequency trade execution model to evaluate the economic impact of supervised machine learners. Extending the concept of a confusion matrix, we present a "trade information matrix" to attribute the expected profit and loss of the high-frequency strategy under execution constraints, such as fill probabilities and position dependent trade rules, to correct and incorrect predictions. We apply the trade execution model and trade information matrix toLevel II E-mini S&P 500 futures history and demonstrate an estimation approach for measuring the sensitivity of the P&L to the error of a recurrent neural network. Our approach directly evaluates the performance sensitivity of a marketmaking strategy to prediction error and augments traditional market simulationbased testing. K E Y W O R D Sexecution model, high frequency trading, market making, recurrent neural networks, supervised learning

show abstract

Sequence classification of the limit order book using recurrent neural networks

Cited by 65 publications

References 36 publications

Feature Engineering for Mid-Price Prediction With Deep Learning

Feature Engineering for Mid-Price Prediction With Deep Learning

Mid-Price Prediction Based on Machine Learning Methods with Technical and Quantitative Indicators

A high‐frequency trade execution model for supervised learning

Contact Info

Product

Resources

About