Across all score functions and aggregation types, recommended course sets have on average higher fitness than the courses students actually took¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Recommended course sets yield a positive average fitness increase over actual course histories, ranging from 2.2% (Quadratic, Specialisation 1) to 21.9% (Logistic, Specialisation 4) in compensatory terms and from 23.1% to 997.2% in partially compensatory terms. → Luis Alberto Pinos Ullauri 2025

Evidence¶

Luis Alberto Pinos Ullauri 2025¶

Luis Alberto Pinos Ullauri, Alexis Lebis, Abir Karami, Mathieu Vermeulen, Anthony Fleury, Wim Van den Noortgate. (2025). A Recommender System of Postgraduate Courses Based on Soft Skills: A Psychometric-inspired Approach. Int J Artif Intell Educ 35. https://doi.org/10.1007/s40593-025-00463-z

q2 · i? · design · r2

Numerical optimisation results summarised in Table 7 from genetic algorithm runs on the test set of students at stage 2. The article reports the average fitness increase of recommendations over actual courses, e.g. "2.2% in Quadratic in Specialisation 1 to 21.9% in Logistic in Specialisation 4". All students in the test set benefited from a higher fitness.

"It can be seen that across all score functions and aggregation types, the average increase in the fitness is positive. This increase varies from 2.2% in Quadratic in Specialisation 1 to 21.9% in Logistic in Specialisation 4 in compensatory terms."

Discussion¶

Related Claims¶

• Brute-force search over all possible course sets is computationally infeasible for specialisations with large course catalogues — related
• Fitness increases vary across the 10 soft skills, with some skills showing small negative increases under compensatory aggregation — related
• Recommended course sets satisfy significantly more soft skill targets than the courses students actually took — related