Trending topics come and go. They are easy to spot – just look at google trends or the “most read” columns in news outlets. But, much more exciting than today’s hot topics would be to be able to predict what everyone will be talking about tomorrow. Of course, this also depends on events which are beyond our predictive power. However, this did not stop us taking a closer look at the possibilities of predictive modelling.
We developed a predictive model, improved it step by step by adding more and different kinds of data and finally tested its predictive power. First, we set up a basic model to predict news trends, using only crawling data. Secondly, we refined it by adding audience measurement data and finally, we added panel data. Illustrated on a case study about media news outlets, our research reveals that panel data is especially valuable for predicting news trends. The in-depth knowledge we gain by adding panel data, thus knowledge about the audience, significantly improved the predictions.
Beyond highlighting the importance of people-centric data, our methodology shows a valuable step towards, for example, newsroom monitoring tools or for simulations of media campaign effects.
For detailed information about this research, take a look at our whitepaper Predicting News Trends.