CloudQuake is a university project where I tried to apply real-time sentiment analysis on Twitter streams. Our approach was to first collect data via Amazon Kinesis (which allows for a convenient way to combine data from different sources) and then use Apache Spark Realtime and its MLlib to classify tweets using a simple Naive Bayes classifier trained with publicly available annotations. While the initial idea was to first filter the data for earthquake related tweets and then combine the inferred tweet sentiments with a Long/Short term moving average indicator with the goal to detect whether the mentions of Quake-related tweets allow to detect earthquakes. The detection didn’t prove to be stable probably owed to the relatively limited amount of tweets that is provided by the Twitters free Api. So all is left is a nice Twitter sentiment analysis engine ;-). The code is available on my GitHub.