Use this URL to cite or link to this record in EThOS:
Title: Hybrid neural network analysis of short-term financial shares trading
Author: Turkedjiev, Emil
ISNI:       0000 0004 7430 0591
Awarding Body: Northumbria University
Current Institution: Northumbria University
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Recent advances in machine intelligence, particularly Artificial Neural Networks (ANNs) and Particle Swarm Optimisation (PSO), have introduced conceptually advanced technologies that can be utilised for financial market share trading analysis. The primary goal of the present research is to model short-term daily trading in Financial Times Stock Exchange 100 Index (FTSE 100) shares to make forecasts with certain levels of confidence and associated risk. The hypothesis to be tested is that financial shares time series contain significant non-linearity and that ANN, either separately or in conjunction with PSO, could be utilised effectively. Validation of the proposed model shows that nonlinear models are likely to be better choices than traditional linear regression for short-term trading. Some periodicity and trend lines were apparent in short- and long-term trading. Experiments showed that a model using an ANN with the Discrete Fourier Transform (DFT) and Discrete Wavelet Transform (DWT) model features performed significantly better than analysis in the time domain. Mathematical analysis of the PSO algorithm from a systemic point of view along with stability analysis was performed to determine the choice of parameters, and a possible proportional, integral and derivative (PID) algorithm extension was recommended. The proposed extension was found to perform better than traditional PSO. Furthermore, a chaotic local search operator and exponentially varying inertia weight factor algorithm considering constraints were proposed that gave better ability to converge to a high quality solution without oscillations. A hybrid example combining an ANN with the PSO forecasting regression model significantly outperformed the original ANN and PSO approaches in accuracy and computational complexity. The evaluation of statistical confidence for the models gave good results, which is encouraging for further experimentation considering model cross-validation for generalisation to show how accurately the predictive models perform in practice.
Supervisor: Busawon, Krishna ; Danaher, Sean ; Aggoune-Bouras, Wohida Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: G900 Others in Mathematical and Computing Sciences ; L100 Economics