Predicting stocks returns correlations based on unstructured data sources

The recent outbreak of information demand for financial investment management forces to look for novel ways of quantitative data analysis. Relying only on traditional data sources means to loose the edge over the competence and become irrelevant in the future. On the other hand, the Big Data and Data Analytics trends are getting traction in financial domain and are being sought as highly beneficial in the long term. This paper presents an approach for forecasting stock correlations based on big volumes of unstructured and noisy data. We evaluate the prediction model and demonstrate its viability for certain industrial sectors.