With the advent of social media concepts such as forecasting and nowcasting became
part of the public debate. Past successes include predicting election results, share price
movement in the stock market and forecasting many other events or behaviors. This
project aims at using social media data, and specifically data from Twitter that are related the songs and artists that currently appear on the highest 10 ranks of the Billboard
Hot 100 chart, perform sentiment analysis on the collected tweets, classify them as positive or negative and finally utilize this information to generate predictions about the
chart of the following weeks. In more detail, the goal is to investigate the relation between the number of mentions of a song and its artist, as well as the semantic orientation of the relevant posts and the performance of the song on the next chart. Firstly, the
problem was approximated through regression analysis, which estimated the difference
between the actual and predicted positions and yielded moderate results. Secondly, the
task was specialized into providing forecasts for some ranges of the chart, namely, for
the top 5, 10 and 20 positions of the chart. According to the values of accuracy and Fscore metrics and compared to previous research, the findings can be deemed as satisfactory, especially for the predictions of the top 20 hits.
Collections
Show Collections