Acceso usuarios
G. Sheelapriya, R. Murugesan
This paper introduces Bayesian regularised radial basis function network (BR-RBFN) as a novel approach to capture the 1-day daily future stock price movement. Naturally, the existence of prolonged volatility and inherent noise leads to market risk and predominantly controls price movement in the market which makes the prediction task more complex. So, it is important to provide a reliable model that overwhelms the prediction complexity and capable of capturing the stock price trend in the financial market. The stochastic nature of this proposed BR-RBFN model which optimally penalises the network complexity and randomly assigns weights to the network which significantly diminishing the plague of network overfitting and overtraining. This fast convergence to its optimum local point also increases the network generalisation. Determining the model efficacy, this proposed model has taken two major stocks for experiment analysis without any need of preprocessing the input data. Thus, the results explore that this proposed model performs well with prediction accuracy approximately 99.5%, compare to other advanced non-linear models chosen for this experiment.
