TITLE:
Development of a Simulated Artificial Neural Network Model for Stock Prediction
AUTHORS:
Njideka Nkemdilim Mbeledogu, Roseline Uzoamaka Paul
KEYWORDS:
Artificial Neural Network, Levenberg Marquardt Back Propagation, Gradient Descent
JOURNAL NAME:
Open Access Library Journal,
Vol.11 No.5,
May
31,
2024
ABSTRACT: Artificial Neural Network (ANN) provides the learning ability for solving complex problems in similitude to the human brain. The research gives a detailed descriptive design of a simulated ANN. It expounded the architectural attributes (the network structure) showing the number and topology of the neurons with their interconnectivity and the neuro-dynamic attributes by employing the Levenberg Marquardt back propagation for the training and Gradient descent for the learning (adjusting the individual weight of the connection links). The Simulink in MATLAB was used to simulate the model and implemented with the stock data. The optimum performance for the number of hidden neurons was tested for 1000 epochs and the (4-50-5-1) structure had the least MSE training time of 1.4%. The performance of the model was evaluated using RMSE and it returned an error rate of approximately 0.0004. This shows the capability of the simulated ANN model in prediction.