Conference Session

DSA Technical Session 1

Collection

2024 ASEE Annual Conference & Exposition

Authors

Ahmad Slim, The University of Arizona; Gregory L. Heileman, The University of Arizona; Husain Al Yusuf, The University of Arizona; Yiming Zhang, The University of Arizona; Asma Wasfi; Mohammad Hayajneh; Bisni Fahad Mon, United Arab Emirates University; Ameer Slim, University of New Mexico

Tagged Topics

Data Science & Analytics Constituent Committee (DSA)

, collaborative filtering, latent treegraphical models, student success, graduation rates, educational data mining1 IntroductionIn our study, we explore analytics focusing on a crucial aspect of student success: the curriculumpathways that lead students toward achieving their learning outcomes and ultimately earningtheir degrees. In the realm of higher education, the role of analytics is increasingly recognizedas a tool for decision-making that enhances student success outcomes. For example, various ini-tiatives have used student demographics and prior academic performance to guide interventionssuch as counseling, mentoring, and tutoring to improve retention and graduation rates 1,2,3 . Ourperspective emphasizes that the core of student academic

Conference Session

DSA Technical Session 4

Collection

2024 ASEE Annual Conference & Exposition

Authors

Fengbo Ma, Northeastern University; Xuemin Jin, Northeastern University

Tagged Topics

Data Science & Analytics Constituent Committee (DSA)

establish baselines that are used for evaluating the reinforcement learningmodels.The dataset is nearly balanced between the two classes and is randomly split into a training set(70%), a development set (15%), and a test set (15%). Once the random split has been initialized,the training set is used as input for both conventional supervised machine learning models and thecreation of the RL environment. The development set is utilized for grid searchinghyperparameters for all models. The test set is exclusively used for reporting the final confusionmatrices to prevent any data leakage. Figure 6. Experiment workflow used in this studyWe adopt the standard performance metric for balanced binary classification problems: accuracy