Comparison of sentiments data extraction and prediction

Abstract

Exponentially increasing amount of information, variety of data forms, growing a number of big data analysis ant prediction tools gives the new opportunities for business, but create needs for right decisions of selection. This study aims to make three-dimensional analysis of data extraction methods, forecasting methods and sentiment indexes. Historical numerical data and textual data from forex news expressed throw the two sentiment indexes are forecasting by econometric methods, Python text analysis and unique computational intelligence tool. Prediction of ensemble of Evolino recurrent neural networks (EERNN) is a distribution of expected values reflecting the probabilities of different states of market sentiments. The results are intended to individual investors needs and gives them opportunity of choice, which depends on what data is available, the accuracy of the prediction, how much time can be taken to make the prediction and that the forecaster has enough skills to use the appropriate IT tools.