The Influence of Sentiment Mining on Twitter based Hit Prediction

Abstract

The aim of this thesis is to investigate the influence of adding semantic features to a twitter classifiers. A potential consequence of this research is, that future classifiers, which are trained on Twitter data, will gain a better performance. The data necessary to do this research was collected using API’s. This data was used to calculate certain features for the classifiers like the number of tweets per day and the average semantic value of those tweets per day. Collecting the data lead to complications, considering the album names which should be considered in further research. Using this data to train two classifiers resulted in a improved performance of the classifier trained with sentiment mining as a feature