A Model for Cryptocurrency Returns based on Sentiment

Kriptovalute su jedna od najbrže rastućih grana tehnologije u zadnjih par godina. Kako su relativno nova tehnologija nameće se pitanje je li moguće izraditi kvalitetne modele povrata kriptovaluta na osnovi raznih faktora. Ovaj rad bavi se faktorom sentimenta. Kao faktor sentimenta uzima se omjer broja spominjanja na društvenoj mreži Twitter i tržišne kapitalizacije. Početna pretpostavka bila je da će ulaganje u kriptovalute s velikim faktorom sentimenta imati pozitivne povrate no ispostavilo se upravo suprotno. U svrhu istraživanja provedeni su postupci slaganja portfelja za dvotjedne( dnevni i dvotjedni povrati) i mjesečne rebalanse( dnevni, 15-dnevni i mjesečni povrati). Iako za niti jedan način gledanja povrata portfelja nisu dobiveni kvalitetni statistički podatci, izabrana su dva načina koja su bila najbliža kvalitetnim rezultatima te su izrađeni modeli povrata kriptovaluta.Cryptocurrencies are one of the fastest growing branches in technology in last few years. As a relatively new technology, the question is whether it is possible to make quality models of cryptocurrency returns based on various factors. This thesis covers a sentiment factor. As a sentiment factor, the ratio of the number of mentioning on social network Twitter and market capitalization was used. First hypothesis was that investing in cryptocurrencies with bigger sentiment factor will have positive returns, but it ended up being exactly the opposite. In the research, portfolio forming procedures were implemented for two week( daily and two week returns) and monthly( daily, 15-day, monthly returns) rebalances. Although neither of the forming procedures passed the statistical test, two procedures that got the results closest to ideal results were chosen and cryptocurrency return models were formed