Fighting churn in virtual labs and digital learning environments is among the top priorities when optimizing learn success. Predicting churn early and taking proactive measurements to prevent it can improve learn success drastically. Often, churn depends on the individual perception of the difficulty level of a learner and hence, adjusting difficulty dynamically is one proactive step towards churn prevention.
However, churn prediction is also very important in other areas. Telecommunication providers have been looking into solution for decades and in particular virtual game developers are highly interested in algorithms that detect players that are about to churn.
The latter example is also highlighted by this year’s game data mining challenge at the CIG conference where one track is dedicated to churn prediction. The data for the challenge has been provided by NCSOFT, one of Korea’s largest game developers. The data consists of events of users playing NCSOFT’s Blade & Sould. Now the task is, to predict which users will return to the game in the following three weeks after an observation period.
As part of the ENVISAGE consortium, goedle.io has taken part in the challenge to benchmark their technology. Additionally, competitions are a great way to share ideas and solutions for difficult real-world problems. Check back in September to read more details on the results of the competition.