Quotes Forecasting Method Based on News Analysis as Part of an Internet Cloud Service

This paper explores the possibility of extracting information from the Internet and predicting changes in stock quotes for selected companies in the stock market by analyzing public news texts over a long period of time. As an input, news headlines and annotations are considered, which, as testing shows, sufficiently reflect the emotional assessment of the news text. Based on previous studies in the field of text analysis and forecasting, a method of binary semantic news evaluations was developed using LSTM, which, as it turned out, was enough to repeat some results of previous work in the field of forecasting stock quotes through machine learning methods, such as: SVM, K-NN, Naive Bayes classifier and semantic analysis methods. In this work, a recurrent neural network was used for prediction. For selected companies, the results of the correct forecast exceeded 80%, thereby showing that research in this area may have an applied prospect for analyzing markets through news information from specialized sources.