self-organizing map neural network

Stock and price index prediction are among the main challenges for market players, traders, and economic analysts. Pattern recognition is one of the most common methods for analyzing complex data such as financial data. Elliot waves are used as one of the most robust models for predicting many markets, and it works based on a hypothesis that argued that upward and downward market price action always showed up in the same repetitive patterns. The need for expert knowledge and skills to detect these waves makes using it difficult for many traders. So far, little research has been done on the automatic identification of these waves. In this paper, we have attempted to recognize these patterns automatically and use them in predicting future upward/downward trends in prices. For this purpose, twelve patterns have been selected as representing Elliot waves. These patterns are stored in a self-organized map neural network and the network is used to identify the waves in the target stock. The proposed algorithm has been tested with several stocks from the Forex financial market. The results have an average accuracy of 93.94 percent in predicting stock trends and it indicates an improvement in prediction accuracy compared to other works.