Optimizing the Duration for Feature Selection in LSTM based Stock Price Prediction
Author(s)
Download Full PDF Pages: 12-20 | Views: 310 | Downloads: 93 | DOI: 10.5281/zenodo.5812642
Volume 10 - November 2021 (11)
Abstract
The increase in availability of data and sophisticated machine learning methods, forecasting stock price has become a major attraction for both data scientists and traders alike. LSTM models are one of the most advanced in terms of determining patterns undetectable to the human minds and are utilized in stock price prediction for the same advantage. This paper evaluates the impact of number of training data points on the intraday returns forecasted using LSTM. Both the returns and the volatility is considered and the results are verified over a large duration and comparisons are made between the sizes of different training spans. High sharpe ratios (>2) were obtained with multiple partition sizes with improved mean intraday returns. The partition size of 50 was found to be the most appropriate for stock price forecasting
Keywords
LSTM, training span, Sharpe Ratio, Mean returns
References
i. Rybalkin, V., Sudarshan, C., Weis, C., Lappas, J., Wehn, N. and Cheng, L., 2020. Efficient Hardware Architectures for 1D-and MD-LSTM Networks. Journal of Signal Processing Systems, 92(11), pp.1219-1245.
ii. Sharma, A., Tiwari, P., Gupta, A. and Garg, P., 2021. Use of LSTM and ARIMAX Algorithms to Analyze Impact of Sentiment Analysis in Stock Market Prediction. In Intelligent Data Communication Technologies and Internet of Things: Proceedings of ICICI 2020 (pp. 377-394). Springer Singapore.
iii. Siami-Namini, S. and Namin, A.S., 2018. Forecasting economics and financial time series: ARIMA vs. LSTM. arXiv preprint arXiv:1803.06386.
iv. Qiu, J., Wang, B. and Zhou, C., 2020. Forecasting stock prices with long-short term memory neural network based on attention mechanism. PloS one, 15(1), p.e0227222.
v. Krauss, C., Do, X. A., & Huck, N. (2017). Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500. European Journal of Operational Research, 259, 689–702.
vi. Krauss, C., Do, X.A. and Huck, N., 2017. Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500. European Journal of Operational Research, 259(2), pp.689-702.
vii. Fischer, T. and Krauss, C., 2018. Deep learning with long short-term memory networks for financial market predictions. European Journal of Operational Research, 270(2), pp.654-669.
viii. Sumon, S.A., Shahria, T., Goni, R., Hasan, N., Almarufuzzaman, A.M. and Rahman, R.M., 2019, April. Violent Crowd Flow Detection Using Deep Learning. In ACIIDS (1) (pp. 613-625).
ix. Cheng M, Cai K, Li M. Rwf-2000: An open large scale video database for violence detection. In2020 25th International Conference on Pattern Recognition (ICPR) 2021 Jan 10 (pp. 4183-4190). IEEE.
x. Das, S. and Mishra, S., 2019. Advanced deep learning framework for stock value prediction. International Journal of Innovative Technology and Exploring Engineering, 8(10), pp.2358-2367.
xi. Ghosh, P., Neufeld, A. and Sahoo, J.K., 2021. Forecasting directional movements of stock prices for intraday trading using LSTM and random forests. Finance Research Letters, p.102280.
Cite this Article: