Forecasting Stock Market Trend using Prototype Generation Classifiers

Currently, stock price forecasting is carried out using either time series prediction methods or trend classifiers. The trend classifiers are designed to predict the behaviour of stock price’s movement. Recently, soft computing methods, like support vector machines, have shown promising results in the realization of this particular problem. In this paper, we apply several prototype generation classifiers to predict the trend of the NASDAQ Composite index. We demonstrate that prototype generation classifiers outperform support vector machines and neural networks considering the hit ratio of correctly predicted trend directions. Key-Words: Stock price, stock market index, forecasting, prediction, learning vector quantization, prototype generation, support vector machines, neural networks, particle swarm optimization.