Predicative model to identify at-risk students
In the O&SR project, the main objective was to develop a predicative model to help spot early school leavers and student retention by measuring student well-being. With inspiration from the FAST project, (Author: Region Zealand), regarding “Retention of Students at Vocational Schools”, with i.a. the aim to develop a predicative model on personal data, school data and qualitative data, to reduce the drop-out rate. Following the predicative model, we developed 4 categories: “Institutional”, “Individual”, “Social” and “Relation” determined by the partnership in compliance with the retention questions created for the students Figure 1. During the process, we changed the category “Relation” to “Onboarding”, as it seemed more accurate and relevant according to the project and the retention questions and what we wanted to measure.
In each of the 4 categories, it is possible to ask a minimum of 4 questions and a maximum of 10 questions with a Likert scale answering format from 1 to 5 per question. If a user has not answered the questionnaire, it will appear on the list with 0 as answer. If there is more than one question per category, the total sum of the answers will be displayed.
A Total column shows the average of all answers, and a colored flag (red, yellow, or green) will be displayed next to it according to the flag percentages configured on the Back office when creating the categories.
A calculation of the average response per category is shown, resulting in a colored flag depending on the value obtained: e.g., red from 0 to 11, yellow from 12 to 14 and green from 15 to 20.
Figure 1: Predicative Model developed in the O&SR project – click to zoom ↑
It is important to notice that special attention should be paid to students having a red or yellow flag, but also to those having a low score in one of the categories. For instance, the last line on Figure 1 shows a green flag student who could be classified as a well-functioning case according to the average score. However, given the extremely low score in the “Relation” category, this student should be kept a close watch on or maybe contacted.
Extraction and analysis of responses from a CSV file is possible in the back office.
In the Statistics (Figure 2) you can extract the answers and keep track of the results by downloading them in an Excel sheet by clicking the “Export data” button before the questionnaire expires (Figure 3).