Prediction of stock data is an inherently difficult process because of its chaotic nature. Experts say that the prediction of these values is impossible, however, research papers show that many of the values depend on the previous ones, therefore they are expressible through machine learning methods. This research used several datasets, including “Yahoo! Finance” stock data, news sentiment analysis training dataset, and a news dataset with related stock and publishing date. First, we worked on sentiment analysis for news. Compared pretrained transformers and random forests for this task. After that, we worked with recurrent neural networks of type LSTM and GRU to predict the stock market price for the next days. We compared RNNs with news and RNNs without news.