This is a code written for bachelor's degree, which entails upon a subject of text mining. In the era when cryptocurrencies were gaining traction, text mining was used to predict cryprotcurrency trends. Originally, the idea was to use twitter, chat and news data, but at the end only news were used.
After tokenization of texts, a matrix was built, which was then fed to SVM (and some other ML algorithms). All code is written in python, with predominantly using libraries like NLTK, scikit-learn and numpy.
Based on our method, results showed that prediction of upward/downward trends is not possible, but overall change in price can be predicted with some accuracy.
For whole text and abstract of bachelor's degree (also in English), visit http://eprints.fri.uni-lj.si/3916/.