CHART PATTERNS EXTRACTION AND RECOGNITION IN CBR SYSTEM FOR FINANCIAL FORECASTING

摘要

Chart patterns identification is a non-trivial task. Since there is hardly a single time scale which will work for all analytical process, investors often find difficulties in identifying the typical chart patterns from a stock prices chart. Discovering the chart patterns is said to be experience demanding and knowledge intensive. Novice investors usually fail to identify the pattern correctly, this make them fa ll into prey in the stock markets. In this paper, an algorithm PXtract is proposed to solve the problem. PXtract dynamically make use of different time windows to analyse and identify any possible chart patterns. A wavelet multi-resolution matching method is also introduced to automate the chart pattern matching process.