Title: Technical analysis forecasting and evaluation of stock markets: the probabilistic recovery neural network approach
Authors: Andreas Maniatopoulos; Alexandros Gazis; Nikolaos Mitianoudis
Addresses: Department of Electrical and Computer Engineering, Democritus University of Thrace, Xanthi, 67100, Greece ' Department of Electrical and Computer Engineering, Democritus University of Thrace, Xanthi, 67100, Greece ' Department of Electrical and Computer Engineering, Democritus University of Thrace, Xanthi, 67100, Greece
Abstract: The market efficiency theory suggests that stock market pricing reflects all publicly available information regarding a given stock. To outperform the market, a shareholder must study the market's price volume patterns and predict human behaviour and tendencies. Except for conventional approaches based on fundamental or technical analysis, new tools are currently proposed using big data and artificial intelligence. This publication analyses and evaluates four commonly used deep-learning artificial neural network models. Then, it proposes a new method by adopting the 'probabilistic recovery' algorithmic approach. The dataset used consists of 501 unique company names based on real data derived from US Dow Jones. This method closely follows the market's behaviour, providing daily upwards-downwards data trends. The proposed system can be used as a tool for technical analysis regarding the prediction accuracy of trading strategies, providing approximately 60% future movements' accuracy, over 90% relative price prediction and annual investment return slightly over 60%.
Keywords: technical analysis; probablistic neural network; neural networks; stock market prediction; stock market forecast; stock market dynamics; stock market neural networks forecasting; algorithmic trading; finance generative adversarial networks; finance convolutional neural networks; CNNs; fully connected neural networks; recurrent neural networks; RNNs; technical indicators; decision making; trading strategies.
International Journal of Economics and Business Research, 2023 Vol.25 No.1, pp.64 - 100
Received: 15 Jan 2021
Accepted: 14 Mar 2021
Published online: 30 Nov 2022 *