Feedback-loop framework coupling a gradient-boosting enrollment classifier with an auto-start simulated-quenching optimizer for multi-objective financial aid allocation¶

Theory · All theories
Evidence · 1 claim (1 for) · 1 study (1 associational), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

The framework treats aid allocation strategy as a searchable solution space of merit and need buckets and amounts. A classifier predicts enrollment and expected outcomes (enrollment, net revenue, unmet need, accessibility, ROI, achievement) from applicant features including promised award amounts; an optimizer evaluates a weighted multi-objective function and proposes revised strategies. The article explains: "The feedback loop is an iterative interaction between the classifier, which predicts expected outcomes, and an optimizer, which evaluates the outcomes and suggests a revised strategy with possibly better outcomes." Unlike fixed-feature settings, award and aid amounts are features updated during optimization. Its novelty is jointly allocating merit-based awards and need-based aid and including affordability and accessibility as objectives.

Design Implications¶

Context¶

Requirements¶

• Requires institutional admission, FAFSA, and student information system data, including aid offer amounts, with only FAFSA filers included for need-based aid eligibility
• Requires a defined neighbor relation over allocation strategies with increasing buckets and amounts and minimum $100 differences between award buckets

Constraints¶

• University administration constrained merit strategies to no more than six awards and need-based strategies to four awards, with maximum merit award four times the maximum need-based amount
• No demographic variables are used so as to avoid bias in recommending aid strategies

Target Learners¶

• domestic first-time freshman applicants to a public university

Target Learning Objectives¶

• increase access and affordability of higher education through optimized allocation of merit-based awards and need-based aid

Claims¶

• Students not on track for college enrollment and persistence can be classified with about 90 percent accuracy using a small set of predictors [+W]

Related Theories¶

Examples¶

-

Key Sources¶

• Vinhthuy Phan, Laura Wright, Bridgette Decent. (2022). Optimizing Financial Aid Allocation to Improve Access and Affordability to Higher Education. Journal of Educational Data Mining, Volume 14, No 3. https://jedm.educationaldatamining.org/index.php/JEDM/article/view/609