Hi Ralf,
- Where can a teacher find the predictions in the course? I expected a link in course administration area or a block, but didn't found anything
- Predictions for a course will appear under Course administration > Reports > Insights once the system starts generating predictions
- Can a teacher define which models he want to use for his course?
- No, only users with 'moodle/analytics:managemodels' capability at site level can (admins, managers or researchers with that capability)
- Can a teacher disable predictions for his course?
- Teachers can disable insights notifications (Preferences > Notification preferences > Insights generated by prediction models) so they will not be notified when new predictions are generated
- Can a teacher add notices to the prediction form?
- Sorry, I don't understand the question
- What happens if the teacher changes prediction related data during the course (start/end date, activities, number of weeks, conditional activities, passing grades)?
- (sorry in advance, long explanations)
- start/end date and number of weeks: Time-splitting methods are what defines at which point of the course the predictions will be generated and there is a scheduled task looking for courses that reach points in time defined by time-splitting mehotds. Without an example this does not help much: Say you use quarters time-splitting method, the course duration is split in 4 parts from the start date to the end date. When we reach the first quarter of an ongoing course we generate predictions. We identify those predictions with a key (1st quarter) predictions for the first quarter will not be generated again (unless you add new students to the course) If you change the course start date or the end date the scheduled task will not get generate predictions for the first quarter again, it will not generate predictions until the 2nd quarter (according to the new start and end) is reached
- Activities: If new activities are added and the date when they are supposed to be completed match the time range provided by the time-splitting method these activities will count towards students' cognitive depth and social breadth calculations
- Conditional activities: Activities that are not accessible by students are not considered
- Passing grades: Same case than Conditional activities if I understand your point correctly. We don't use passing grades as part of students at risk model. If what you ask is what happens if I change the passing grade after the course is complete? First thing to mention is that teachers will not get predictions for finished courses. Finished courses are used to train the machine learning algorithms that will later be used to generate predictions. Second thing to mention is what this always depend on what is defined by the model; replying what I imagine you could mean (models using the passing grade as a success / fail) It depends on when you change the pass grade
There is a scheduled task that will use finished courses data for training, no manual action is required. If you change the course pass grade after the course end date nothing will happen because that course will already be used for training. If you change the passing grade before the end date that is the value that will be considered.
- Is the learning time of a student included in the calculations? How is learning time defined and calculated?
- If by learning time you mean the time a student spent in a course no, it is not included. This should be a very good indicator and a whole topic by itself. There are multiple ways of calculating this, none of them really reliable unless we constantly inform the server about whether the user is actively viewing the page or similar solutions. Also a time spent of 45h does not really mean anything; we need models to adapt to all sort of courses so we need to put that value in a context; if we want a time spent indicator we can:
- Do it properly by comparing the time spent value we calculate with an "Expected time spent" or something like that that the teacher would have to set. We will need to define at what level (whole course, activities, course sections...)
- Add a simple one e.g. zero time spent / less than course average time spent / more than average or similar solutions; quite less powerful and reusable than the previous one
- If by learning time you mean the time a student spent in a course no, it is not included. This should be a very good indicator and a whole topic by itself. There are multiple ways of calculating this, none of them really reliable unless we constantly inform the server about whether the user is actively viewing the page or similar solutions. Also a time spent of 45h does not really mean anything; we need models to adapt to all sort of courses so we need to put that value in a context; if we want a time spent indicator we can:
