High-performance stock index trading: making effective use of a deep LSTM neural network

Abstract

We present a deep long short-term memory (LSTM)-based neural network for predicting asset prices, together with a successful trading strategy for generating profits based on the model's predictions. Our work is motivated by the fact that the effectiveness of any prediction model is inherently coupled to the trading strategy it is used with, and vise versa. This highlights the difficulty in developing models and strategies which are jointly optimal, but also points to avenues of investigation which are broader than prevailing approaches. Our LSTM model is structurally simple and generates predictions based on price observations over a modest number of past trading days. The model's architecture is tuned to promote profitability, as opposed to accuracy, under a strategy that does not trade simply based on whether the price is predicted to rise or fall, but rather takes advantage of the distribution of predicted returns, and the fact that a prediction's position within that distribution carries useful information about the expected profitability of a trade. The proposed model and trading strategy were tested on the S&P 500, Dow Jones Industrial Average (DJIA), NASDAQ and Russel 2000 stock indices, and achieved cumulative returns of 340%, 185%, 371% and 360%, respectively, over 2010-2018, far outperforming the benchmark buy-and-hold strategy as well as other recent efforts.

Selecting Datasets for Evaluating an Enhanced Deep Learning Framework

“…Some of these algorithms were based on gated recurrent neural networks (RNN), autoencoders, convolutional neural networks, bidirectional mechanisms, attention mechanisms, ensemble techniques, deep and vanilla architectures. Specific architectural design features of these 12 selected algorithms were: the gated LSTM architecture suggested [13], [14] and [15]; the bidirectional mechanism combined with both LSTMs and GRUs influenced [16]; the attention mechanism combined with gated neural networks [17]; deep convolutional neural network (CNN) ensemble with LSTM and an attention mechanism [18]; a GRU [19] autoencoders combined with LSTM [20] and finally a deep gated recurrent neural network architectures made up of both GRU and LSTM [21]. iii Evaluation-The following factors were considered as potential performance evaluation criteria with specific metrics: complexity measure through the total number of built parameters of every architecture, accuracy considered a mean absolute error (MAE) which is robust in environments associated with discrete irregular patterns when measuring the average magnitude of the errors in a set of predictions, without considering their direction.…”

“…On the other hand, models and algorithms designed through gated sequential architectures in the form of LSTMs and GRUs have been widely used in such analysis environments [13], [14], [15]. Thus the guidance from the SeLFISA framework will influence the development of deep learning with artefacts that may demonstrate better performance over these suggested gated models.…”

Optimization of Traditional Stock Market Strategies Using the LSTM Hybrid Approach

“…This model uses an attention mechanism to capture long-term dependencies and interactions among features and perform multiple time series forecasting. The research focuses, in particular, on applying the Transformer model in portfolio management, which is used for trend-following strategies and volatility forecasting in multi-period portfolio optimization [29]. Although the Transformer model shows strong results in forecasting time series, the research also highlights the challenges of applying machine learning in finance, such as maintaining a balance between model complexity, tuning quality, and forecast quality.…”

Estimating value at risk: LSTM vs. GARCH

“…We refer to Greff et al (2016) for an account on the history of LSTM and its various specifications. Other applications may be found in the work of Chalvatzisa and Hristu-Varsakelis (2019), Wu and Yan (2019), and Sirignano and Cont (2019).…”