Topology representing networks
Neural Networks
Self-organizing maps
ICANN '97 Proceedings of the 7th International Conference on Artificial Neural Networks
| Hi-index | 0.00 |
Financial markets provide a singular field of analysis and exciting challenges for researchers. During the past decades, strongest assumptions on financial time-series (namely the Random Walk Hypothesis) have been partially discharged, and useful new paradigms (such as chaos, self similarity, self-organized critically) have been discovered to this field. One of teh arguments with respect to random walk is that chaotic time series can be long term random, and still have short term patterns. If, in fact, the marketplace has short term patterns, and some technicicans think it does, then there are windows of opportunity tat the random walker either denies exist or are unable to be exploited. Self Organizing Maps (SOMs), therefore, offer a powerful tool, suitable to explore financial markets, in order to find significant patterns, and use them as a forecasting tool. Additionally, the possibility to detect features of the market can be employed to find out the statistics of the process under examination, hence to decide whether or not the Random Walk Hypothesis is acceptable. This chapter will explore the use of Self Organizing Maps for such purpose. Both the conditions needed for data to be treated, and the evaluation of results obtained are considered.