Consolidating financial outlook of companies

Abstract

Being able to determine the financial outlook of a company is imperative for one to look for investment opportunities. Three areas are identified which enable investors to better understand the financial outlook of a company: financial statements, financial news, and financial reports. As financial statements refer mainly to financial accounting information, critical information can be displayed in a form of a dashboard. The business information, activities and operations are mainly presented in the text found in 10-K reports. Extractive text summary methodology called TextRank with Global Vectors for word representations (GloVe) would be used to find key sentences in the report. Lastly, we would use sentiment classification on news articles to better understand and consolidate the views of articles that may affect the public’s opinion of the company. For sentiment classification, Vanilla RNN, LSTM, bidirectional LSTM, and different forms of artificial Recurrent Neural Network (RNN) architecture would be explored. Transfer learning would be applied with word embeddings by Word2Vec. The model with the best accuracy would be used for the sentiment classification of company news. The outputs are consolidated in a web application using React. This web application would display the dashboard, the summary of financial reports, and the sentiment of news articles for each company.