It’s indisputable that sentiment is a huge driver in the cryptocurrency market, and yet it is one of the hardest metrics to measure.
In this blog post I’m going to talk about my learnings in developing some sentiment based analysis systems to try and find predictors for cryptocurrency price.
Sentiment analysis has been a hot topic in recent years and there are now many packages which make performing it a much quicker task. There are also some great data sources available — in this article I first use Reddit’s API, and then CryptoCompare’s news API . We’ll first show why looking at mood in social media posts might not be the best idea, and then we’ll analyse how interest in specific topics changes over time using Latent Dirichlet Allocation.