We scrape the news data using webhose.io and stock price data using NSEPY (for Indian stock market). The data is labelled using information about open and close prices corresponding to stock news for that day. For the word embeddings, we used GloVe provided by Stanford University. We used stop words from NLTK to remove sentence fillers that do not change the context. Then we used TF-IDF to remove the words that provide the least information. LSTM is used to map any time-dependent relations in the data set and predicted if a user should "buy" or "sell" the share and with what confidence based on today's news events.
atharvabhave21/Stock_trend_prediction_using_setiment_analysis_of_news_data.
I scrape the news data using webhose.io and stock price data using nsepy. then I labelled the data using information about rising or fall in prices corresponding to news scraped for that day. For the word embeddings, I used GloVe provided by Stanford University. I used stop words from NLTK to remove sentence fillers that do not change the context. Then I used TF-IDF to remove the words that provide the least information. then I used LSTM to map any time-dependent relations in the data set and predicted if a user should "buy" or "sell" the share and with what confidence based on today's news events.
Jupyter Notebook