Long short-term memory network for high-frequency trading: A critical analysis based on an Italian case study
Jacopo Chiapparino & Pier Giuseppe Giribone
What the paper says
Deep Learning technologies have proven to be capable of making predictions in High-Frequency Trading, often reaching better results compared to other techniques traditionally used for this purpose. A Long Short-Term Memory (LSTM) architecture has been implemented because, despite noise, nonlinearity, nonstationarity and complexities in high-frequency data, this Recurrent Neural Network (RNN) deep learning technique allows for the discovery of interesting patterns, due to its peculiar features. Unlike traditional RNNs, which struggle to retain information over extended sequences, LSTM cells are able to selectively store and retrieve significant information from the past values. The goal of this paper is to study and implement a robust process for the fine-tuning of these nonlinear autoregressive models, analyze the results with the proper metrics, and evaluate the overall performance. A market case study based on an Italian stock has been analyzed to complement our discussion and we show how to improve the network performance using the technical indicator Williams %R as an exogenous variable related to volumes.
Evidence weight
Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40
| F · citation impact | 0.50 × 0.4 = 0.20 |
| M · momentum | 0.50 × 0.15 = 0.07 |
| V · venue signal | 0.50 × 0.05 = 0.03 |
| R · text relevance † | 0.50 × 0.4 = 0.20 |
† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.