joint prediction of stock price/correlation pair using deep multi-task networks

Stock price prediction is a great challenge due to the volatile and uncertain nature of the market. the correlation coefficient is a crucial issue in portfolio selection which depends on price history. our aim here is to build a model that is capable of predicting correlation coefficient and price movement of stocks at the same time. to this end, we use the multi-task learning (mtl) framework. the mtl model learns multiple tasks in parallel to make more accurate predictions [1]. the raw data used in this study is the adjusted closing price of 30 companies listed in tehran stock exchange (tse). experimental results confirm that the proposed model performs well in predicting the price/correlation pair.