Abstract:
The stock market is a place, be it virtual or real, where large groups of people gather
to buy and sell stocks. The market creates an avenue for companies to seek capital for their various needs and also make profits. Thus, prediction of stock prices is regarded as a challenging task, as an accurate prediction of stock prices may yield profits for investors.
However, the stock market is rather a complicated system, and due to the complexity of stock market data, development of efficient models for predicting is quite challenging. Artificial Neural Networks (ANNs) have become very important method for stock market predictions because of their ability to deal with uncertain, fuzzy, or insufficient data which fluctuate rapidly in very short periods of time. With ability to discover patterns in nonlinear and chaotic systems, ANNs offer the ability to predict market directions more accurately than traditional techniques such as technical analysis, fundamental analysis, and regression. This study therefore developed two efficient neural network models and compared their performances in predicting the daily prices of stocks in the Nigerian Stock Market. This study utilised three-layer (one hidden layer) multilayer perceptron models (feed forward neural network models), as these models are mathematically proved to be universal approximator for any function. The basic input data includes: raw data such as the daily open, high, low, average, previous and close prices, which form the technical variables. Computation within the layers was done using the tangent sigmoid function. Gradient descent back propagation algorithm was also used to ensure that the error between the actual and predicted output is minimized. The network was simulated using C#.net as the programming language. The results of this
study were compared with the Moving Average using relative error and mean squared error. The result comparison proved that neural networks outperformed the traditional methods of predicting stock prices.
