Deep Language Classification for Relabeling of Financial News and its application in Stock Price Forecasting
published: Nov. 14, 2019, recorded: October 2019, views: 13
Report a problem or upload filesIf you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.
This paper aims at assessing the performance of the transfer learning task consisting of training set of classiffiers on high frequency financial news data for 74 publicly traded companies, with domain speciffic labels. This source of data is provided by the Jozef Stefan Institute and is used exclusively for the purposes of this research. The trained classiffiers are then used to attribute labels to an unlabelled source of high frequency aggregated news, Event-Registry. The aim is for the relabelled data to be used in the generation of exogenous features for use in time series forecasting of the companies' prices. It is found that using a fine-tuned BERT  model yields the most semantically coherent labels, and the features generated from the newly labelled data prove to yield the highest accuracy forecasts on held out price data.
Download slides: sikdd2019_torkar_stock_price_forecasting_01.pdf (2.1 MB)
Link this pageWould you like to put a link to this lecture on your homepage?
Go ahead! Copy the HTML snippet !