I open this discussion to ask if there was the possibility of having the precise model that allows us to have the probability that a student will drop out of the course of study . Specifically, I would like, eventually, to understand the scrpit that allows you to consider the different variables and the weights attributed to them.
I know that moodle allows for this but I don't know if it's possible to know the model working behind that percentage.
Thank you so much for your help.
That model is pretty well documented here:
At a high level, each core Moodle activity provides a social and cognitive indicator. Moodle uses a target of "Enroled students show no activity in the final quarter of the course." and passes those users to the model with their associated indicators to train the Model - eg "these are the users that show behaviour we want to predict"
Then it collects that data for your current users for each time-frame analysis and sends those indicators to the Machine learning engine and at the core model returns a boolean response for users that match the trained indicators.
If you're a developer you might want to look at some specific indicators to see how they generate those values in the code - take a look in the activity/classes/indicator folders - eg assign/classes/indicator for the code behind each indicator.
hope that's useful!
But is it not possible to have exactly the model used to include all the indicators considered?
The procedure is quite clear, however I really would to have the precise model used to define belonging to a certain response class.
I hope for your later help!
Course accessed after end date
Course accessed before start date
Any write action in the course
Read actions amount
IMS pkg cognitive
IMS pkg social
Completion tracking enabled
Course potential cognitive depth
Course potential social breadth