Exportar Publicação

Mendes, D. A., Ferreira, N. B. & Mendes, V. (2019). Could the supply of a chain big data analytics market register a better forecast performance for the Stock Markets? – A comparative software analysis. ITISE 2019.

D. E. Mendes et al.,  "Could the supply of a chain big data analytics market register a better forecast performance for the Stock Markets? – A comparative software analysis", in ITISE 2019, 2019

@misc{mendes2019_1734886185755,
	author = "Mendes, D. A. and Ferreira, N. B. and Mendes, V.",
	title = "Could the supply of a chain big data analytics market register a better forecast performance for the Stock Markets? – A comparative software analysis",
	year = "2019",
	url = "http://itise.ugr.es/"
}

TY  - CPAPER
TI  - Could the supply of a chain big data analytics market register a better forecast performance for the Stock Markets? – A comparative software analysis
T2  - ITISE 2019
AU  - Mendes, D. A.
AU  - Ferreira, N. B.
AU  - Mendes, V.
PY  - 2019
UR  - http://itise.ugr.es/
AB  - The dimension of the finance industry that has been most influenced by technological advances in the last decade it is the speed and frequency with which financial transactions are decided and executed. Data analytics, business intelligence, machine learning, algorithmic trading are the leading tech trends of this decade and the most active and innovative segments of the information technology market. In consequence, we can assist in a quick development process to explore new methodologies and to extract and analyse the rich information that the big data market sets contain. For example, in 2019, the biggest U.S. enterprises tend to migrate their data to hybrid cloud solutions helping agents to centralise management of big-data assets distributed between private and public clouds. 
In this paper, we use the G7 indexes stock markets prices for periods between 10 and 50 years of daily historical data (from Yahoo Finance). The purpose of the analysis it is twofold: first - several forecasting methods are applied, and we search for the minimum forecasting error, second – we are interested in the time duration and the software velocity attained in the forecasting process.
For the first purpose, we use supervised deep learning methods, namely Recurrent Neural Networks and  Long Short-Term Memory (LSTM) architectures.  We have found that the LSTM configuration works the best out of all the combinations we have tried (for our dataset). LSTMs are very powerful in sequence prediction problems because they can store past information, which is crucial in predicting its future price. The forecast error is measured by the Mean Absolute Percent Error and by the Mean Square Error. We use Python software, where Keras, TensorFlow and Pandas are the packages with the main role. 
The second important point that we analyse in this article it is the execution time. We use different software, namely R, Julia and Python, and we aim to measure the trade-off between the algorithm complexity and the speed of execution. A comparison of the below results with ARIMA models forecast accuracy and execution time it is also considered.

ER  -