Stock Forecasting Using Local Data

Stock price forecasting is a relevant and challenging problem that has attracted lot of interest from engineers scientists. In this paper we apply two techniques for stock intervals forecasting. Both techniques, derived previous works by the authors, are based on use local data extracted database. These those correspond to similar market states current one. The first technique uses these compute forecast finding an optimal combination past equals state. then obtained combining actual prices associated states. second can be used but its main will contain real future with guaranteed probability. This accomplished building probability distribution forecasted setting choice desired percentiles. Thus, in financial risk management. purely driven do not need theoretical description or model trend being forecasted. proposed adapt very easily changes because they only subset database it closer Furthermore, updated as new available. Finally, both approaches highly parallelizable, thus making possible manage large sets. As case study, have been applied k-step Dow Jones Industrial Average index. results validated relation some baseline approaches, such martingale neural network predictors quantile regression interval