The project aimed to predict daily closing and hourly stock prices of various stocks listed on the Pakistan Stock Exchange thereby being a source of timely guidance for investors
System Features: Live data processing and technical analysis, User profile and stats and Scalable to other stocks
Technologies/language involved: Python, Hadoop, D3.js, MySQL, Ubuntu, TensorFlow
For scraping: Goose, BeautifulSoup4
For NLP news analysis: nltk(stemming, lemmatization, stopwords removal etc.) and Pysentiment(Python financial news sentiment library)
For prediction: Univariate Long-Short Term Memory (LSTM) for hourly predictions and indicators-based own-built algorithm for daily closing price predictions
Data sources: Quandl, investing.com, nccpl and misc websites
News source: Dawn.com, Ksebusiness.com, Ksenational.com, RSS feeds
Prediction done for HBL, UBL, ENGRO FERTILIZER, PSO and OGDCL with 55 to 60% accuracy.
Credits: To my FYP team mates, Hamza Jamshaid and Majid Ali Khan.