Financial time series forecasting (Stock prediction)

Abstract

Accurate prediction of stock price trend greatly helps stock investor to react correctly in the stock market. The unsteadiness of the stock market has caused serious profit loss to many people. Stock markets are easily affected by many factors. It includes financial, political and unknown company development. In order for one to make profit from the stock market, it needs adequate forecast to plan the future. Hence, effective, stable and accurate methods which able to build a model to have the ability to predict the stock market trend are needed.

The dissertation aims to provide an analysis of Neural Network (NN) and Support Vector Machine (SVM) method to build a prediction model by using Matlab software with the input data of Singapore Technology (ST) engineering company stock price. By using the two methods mentioned to determine the Absolute Error (AE) between predicted stock price value and the actual stock price value and hence to find the Mean Square Error (MSE), the results suggest that SVM method has outperformed NN method on the ST stock price trend prediction.