Volume : VII, Issue : IX, September - 2017

Financial Time Series Prediction using Multi–Scale Neuro Fuzzy Function Approximation

P. Arumugam, R. Saranya

Abstract :

 This paper presents a novel forecasting method for financial time series using artificial neural networks. Financial time series are characterized by chaos and are difficult to forecast by traditional function approximation techniques. They exhibit different cyclical patterns with different time periods. The exponential moving averages of higher time are easier to predict accurately. The higher time period moving averages show smooth patterns that can be captured by nonlinear function approximation by fuzzy neural networks. Although several prediction techniques have been applied to the task of financial time series forecasting, the naive prediction of no change is difficult to beat. The goal is to predict the future values with more accuracy than the naive prediction and also provide upper and lower bounds with 99% confidence. Experimental results on five stock indices show that the proposed method exhibits higher accuracy and reliability.

Keywords :

Article: Download PDF   DOI : 10.36106/ijar  

Cite This Article:

P. Arumugam, R. Saranya, Financial Time Series Prediction using Multi–Scale Neuro Fuzzy Function Approximation, INDIAN JOURNAL OF APPLIED RESEARCH : Volume-7 | Issue-9 | September-2017

Number of Downloads : 228

References :