Analysis of Financial Time Series Data Using Adaptive Neuro Fuzzy Inference System (ANFIS)

The aim of this research is to analyze ANFIS performance for prediction of financial time series data. Financial time series data is usually characterized by volatility clustering, persistence, and leptokurtic data behavior. The financial time series data are usually non-stationary and non-linear. ARIMA has a good performance to predict linear time series data, but its performance is decreasing when applied to predict non-linear times series data. ANFIS as one of hybrid models which composes neural network (NN) and fuzzy system is expected to be able to predict the financial time series data more accurate. Many research conclude that the effectiveness of ANFIS depend on the input selection, the membership function (MF) selection and the rule generation. In this study, the input variables of ANFIS are selected based on preprocessing of original data by using Subset ARIMA model. The rule bases of ANFIS are generated based on a linear Sugeno fuzzy model. The consequent parameters are identified by using least square method and premise parameters are adapted by using gradient descent. For practical assessment of ANFIS performance, ANFIS method is implemented to analyze the Indonesia inflation monthly data from January 1970 up to December 2012.