PATTERN RECOGNITION MODEL FOR NON LINEAR TIME SERIES ANALYSIS IN FINANCIAL MARKETS

摘要

Time dependent data analysis is a potentially challenging task for non linear systems. Though, abundance of historical data has enabled- non linear time series analysis, the performance of these models is not very efficient due to high dimensionality and presence of noise.. Hence two Time Series Representations by name Hybrid Dimensionality Reduction(HDR), Extended HDR(EHDR) and High Low Non-overlapping (HLN) clustering algorithm are introduced in this invention, that produces efficient results. HDR and EHDR are concerned with reduction of noise and dimensionality. It meaningfully discretizes the non linear time series data that enables to understand/analyze the past behavior/pattern and forecast the future events/behavior of the system. HLN is used to group similar non linear time series datasets arid helps in efficiently forecasting the behavior of non linear systems. The current invention improves the accuracy and efficiency of the decision making ability in non linear systems like financial markets, electricity market, etc.,