Students at risk models are not discarding user enrolments whose start and end dates do not fit into the analysed time interval

We discard finished courses and courses whose start date is in the future, but we are not discarding enrolments whose timestart and timeend do not fit into the calculated times. This results in some problems:

• Predictions are generated for student enrolments whose start is in the future.
• Predictions are generated for student enrolments whose start date is after the analysis interval. This can result in inaccurate indicator calculations.
• Predictions are generated for finished student enrolments.