Stock Indices Forecasting based on Wavelet Filters and Improved Instance based Learning (WIIBL)

The proposed model implements a unique technique that extends the nearest neighbor rule to incorporate the idea of pattern matching to identify similar instances thereby implementing a non-parametric regression approach. A hybrid distance measure combining statistical correlation and Euclidean distance has been incorporated in the model to select similar instances. To illustrate the performance and effectiveness of the proposed model, simulations using data sets of CBOE S&P Buy Write Index, 95-110 Collar Index and Gold Volatility Index has been carried out. Here, we apply a comprehensive set of non-redundant orthogonal wavelet transforms for individual wavelet sub band to denoise the signal. A thorough analysis of model simulations demonstrate that the proposed wavelet based-IIBL model ends up in accurate predictions and encouraging results