Picture from: https://sludgefeed.com/top-cryptocurrency-experts-follow-twitter-2018/
We are glad to add a new feature to our Analysis: Sentiment Analysis Towards the feeling of twitter users about Bitcoin.
What does that mean, ask you might?
Simple: we are analyzing all the tweets from yesterday (for now only tweets in English) by applying a series of text techniques in order to discover what is the feeling about Bitcoin.
From a technical perspective, here is a list of all the procedures we take in order to achieve our goal (all using Python):
- Each day, we crawl, with the help of Twitter’s API, all the tweets related to keyword “Bitcoin” that are written in English
- We apply a series of cleansing text techniques in order to get rid of elements that we won’t need like punctuation, “@”, links, “RT”, stop words
- After we obtain the cleanse phrases, we use the Natural Language Processing package (NLTK) in order to process and analyze them
- The end result will be classifying each tweet into Positive, Neutral or Negative and output the proportion of each from the total number of tweets
- Process is automated so that each day, you can see how the sentiment has evolved based on recent events
Take a look at how the results are displayed:
Or have a look on our page
and come back each day to see the evolution.
Through this analysis, we would like to add a new layer to foreseeing the future values of your favorite cryptocurrencies. Besides the forecast analysis that we provide, we also have sentiment analysis of tweets that will help into decision making. If there will be events like bans, regulations or other negative impacts that might drive cryptocurrency values down, we are confident that this method will capture how the crowd feels and complete the shortcomings of algorithm predictions. As we know, it is difficult for an algorithm to predict events that have a low occurrence or unexplained events and that is just the reason why this technique helps us into having a better picture about the future.
We can easily repeat this analysis for any kind of coins or tokens, however the API is build in such a way that it will not allow us to continuously crawl tweets and we have to put time stops in the code in order to wait a period and then resume. This will obviously increase the time needed to run the analysis.
For the future, we plan to study other cryptocurrencies and build on demand analysis for the ones that are your favorites. Also, there is room for improvement from the language processing part and even extend to other languages.
What do you think? Did you like our analysis? Do you think it will be useful?
Let us know in the comments bellow and be sure to Like and Share so that Bitcoin won’t be so hard to predict.