Deep Learning for Predicting Charts using Social Sensors
Musical charts are traditionally released on a weekly basis. For each track, we can hence model the track's charts performance as a time series (e.g., for the Billboard Hot 100 charts). In this master thesis, we are interested in predicting future chart ranks for a set of tracks. Therefore, we rely on time series models and also incorporate social sensors such as tweets about a given track or last.fm scrobbles. For the computation of predictions we aim to experiment with deep learning-based methods for time series prediction.