VANFIS: virtual adaptive neuro-fuzzy inference system for modelling and forecasting stock data

Random fluctuations take place in stock data. At fluctuation points, it is difficult to predict the next data points from the previous data. Close enough data points are more informative to the forecasting process but not available adequately. This article presents deterministic and stochastic methods for exploration of virtual data positions from actual data and incorporated to original dataset. An adaptive neuro-fuzzy inference system (ANFIS) is exposed to such virtual data positions and works in a virtual environment. The model is termed as VANFIS, i.e., virtual ANFIS and employed to infer future stock indices of real stock markets. The performance of VANFIS methods are validated using 15 years data from ten stock markets and using five metrics. Also, relative worth of proposed methods are carried out over ANFIS. Experimental results show the significant improvement in prediction accuracy when proposed methods adopted.