LDA projection visualizations show knowledge-state features separate STEM and non-STEM classes better than student-profile features.¶

Claim · All claims
Evidence · 1 study · associational r1 · q2
Currency · newest study 2018 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? LDA-projected histograms show the student profile largely overlaps between classes, while DKT or DKT+ knowledge states result in a better separation between STEM and non-STEM classes. → Yeung 2018

Evidence¶

Yeung 2018¶

Yeung, C.-K., Lin, Z., Yang, K., & Yeung, D.-Y. (2018). Incorporating Features Learned by an Enhanced Deep Knowledge Tracing Model for STEM/Non-STEM Job Prediction. arXiv preprint arXiv:1806.03256. https://arxiv.org/abs/1806.03256

q2 · i? · associational · r1

Supervised LDA projection of student profile and DKT/DKT+ knowledge states into one dimension, visualized as histograms in Figure 2 (Section 5.2.1). The authors report "using the knowledge state obtained in the DKT or DKT+ models results in a better separation between STEM and non-STEM classes".

"On the other hand, using the knowledge state obtained in the DKT or DKT+ models re-sults in a better separation between STEM and non-STEM classes, implying that the student’s knowledge state in math-ematics provides a rudimentary differentiation on whether a student is likely to pursue a career in the STEM field."

Discussion¶

Related Claims¶

• Models trained with combined DKT/DKT+ knowledge-state and student-profile features generally outperform models trained with student profile alone on STEM/non-STEM job prediction. — related
• DKT+ knowledge states discriminate STEM from non-STEM students better than DKT knowledge states for the majority of mathematical skills. — related
• Recursive feature elimination mitigates overfitting and improves STEM/non-STEM prediction performance, with LR-DKT&SP performing best after RFE. — related
• Topic-preference features separate grade groups better than Reading Time in all analyzed cases — related