We are particularly interested in their Moodle activity for a few reasons. The first is that it is a very strong predictor of student success. From viewing the Moodle access logs on a week by week basis, we can see that for many modules there is a strong periodicity associated with the modules, as seen in fig 1. The second is that we can make predictions for students mid-way through the semester. This allows us to alert students earlier that they are at risk of failing. Finally, it is perhaps more ethical to base our interventions on the actions of students rather than demographic information that they have no control over.
The figure above shows the Moodle activity over a five year period. While the intensity of the peaks increases over the years, due to increased adoption of the VLE by staff members, the periodicity and repeatability of the data is clear to see. Upon close examination you can see the various peaks and troughs happen around the same time every year e.g. low activity during holiday periods, high activity close to exam time.
We used this data to determine if there was some correlation between VLE engagement and exam performance. Where correlation was confirmed, algorithms were generated to use this data as a predictive tool to predict a current student’s performance in their final exam based on their engagement with the VLE. For each module, we divided the Moodle logs into chunks of weeks. We then extracted multiple features for each week:
• A count of how many times they accessed Moodle
• The ratio of on campus to off campus accesses using IP address
• The average time of day they used Moodle
• How many times times they accessed Moodle during the weekend