Educational institutions are increasingly seeking data driven approaches to identify students at risk and enable timely academic interventions. Traditional learning analytics systems embedded into learning managements systems lack predictive capabilities and real time integration,limiting their effectiveness for proactive student support strategies. This study presents a machine learning-integrated learning analytics dashboard that employs XGBoost algorithm for predicting student final grades using real time Canvas LMS data. The system extracts features from assignment group performance categories (homework, quizzes, exams, participation) through automated Canvas API integration. Data preprocessing involves assignment group normalization and temporal filtering based on course completion percentage. The XGBoost model utilizes gradient boosting to learn complex patterns from structured educational data, enabling progressive prediction accuracy improvement from 57.4% at 20% course completion to 82.5% at 50% completion. The dashboard integrates Django based web framework with PostgreSQL database storing Canvas course structures, student enrollments, assignment metadata, and grade records. Real time data synchronization enables dynamic feature extraction by assignment categories. Evaluation on CS2 course data demonstrates superior performance compared to Linear Regression (70.6%) and Random Forest (76.9%) baselines. The integrated system successfully identifies at risk students 4-8 weeks before final grades, enabling targeted academic interventions.

Background: The rise of Large Language Models (LLMs) in software development has opened new possibilities for code generation. Despite the widespread use of this technology, it remains unclear how well LLMs generate code solutions in terms of software quality and how they compare to humanwritten code. Aims: This study compares the internal quality attributes of LLM-generated and human-written code. Method: Our empirical study integrates datasets of coding tasks, three LLM configurations (zero-shot, few-shot, and fine-tuning), and SonarQube to assess software quality. The dataset comprises Python code solutions across three difficulty levels: introductory, interview, and competition. We analyzed key code quality metrics, including maintainability and reliability, and the estimated effort required to resolve code issues. Results: Our analysis shows that LLM-generated code has fewer bugs and requires less effort to fix them overall. Interestingly, fine-tuned models reduced the prevalence of high-severity issues, such as blocker and critical bugs, and shifted them to lower-severity categories, but decreased the model's performance. In competition-level problems, the LLM solutions sometimes introduce structural issues that are not present in human-written code. Conclusion: Our findings provide valuable insights into the quality of LLM-generated code; however, the introduction of critical issues in more complex scenarios highlights the need for a systematic evaluation and validation of LLM solutions. Our work deepens the understanding of the strengths and limitations of LLMs for code generation.

Nowadays, employers highly value soft skills, yet many students lack these fundamental abilities. Teaching soft skills involves fostering active student participation and facilitating communication of technical knowledge among peers. This approach presents challenges: (i) creating an engaging learning environment; (ii) ensuring students get timely feedback; (iii) finding an approach that is not too time-consuming for instructors to prepare.

This innovative practice full paper describes an experience of using MILAGE LEARN+ with Computer Science university students. Students motivation when in an academic setting is a very important aspect of how well they will learn the material provided. Many computer science students own smart phones, tablets, and computers in order to complete their work and study. Here we introduce MILAGE LEARN+ to fourth year computer science university students. MILAGE LEARN+ is a mobile educational application where students take quizzes, watch videos, and do assignments in worksheet format. This application integrates the gamification pedagogy with the usage of difficulty levels, a leader board, as well as self and peer-review in order to benefit students' motivation, autonomy, and their grades. While MILAGE LEARN+ has been used in many European countries with a variety of age groups and fields of study there has been no research done on how American students connect with the application. In this study we are examining how university students in computer science react to the usage of this application. MILAGE LEARN+ was integrated into a 400-level computer science course at a Research 1, land-grant university, for the usage of quiz taking. We have focused on usage of the leader board, peer-review, and self-review to answer the following questions: How does MILAGE LEARN+ affect students' motivation? Does it benefit their grades? And is it an overall positive experience for our students? Participation in this study was entirely voluntary and those who chose not to participate were given an alternative assignment using Canvas quizzes. The study was reviewed and approved by the IRB committee of our University. There were 38 students participating in this study. All of these students are majoring in computer science or engineering equivalent course. Students had four weeks to take a series of ten quizzes. These quizzes covered content covered in class and were worth ten points on each quiz. Most of the quizzes had five questions worth two points each with a select few having ten questions worth one point each. The quizzes contained a variety of answering styles including: multiple choice, true/false, and free response through the keyboard. We compared these quizzes to an exam on Canvas LMS. The Canvas exam had all of the same questions as provided in the MILAGE LEARN+ quizzes. These response styles were also a mix of multiple choice, true/false, and free response through the keyboard. Student users later evaluated the application and quizzes through electronic survey. The survey included 15 questions, 12 of which were answers based on the Likert-type scale. MILAGE LEARN+ was shown to have a higher percentage of students get a question correct on the quizzes in comparison to those using the Canvas learning management system. Though, MILAGE LEARN+ students were also shown to have issues with the application. While there were some limitations with our study, our results were statistically significant and shown that MILAGE LEARN+ does in fact boost students' grades, even if they encounter difficulties with using the application.

This innovative practice WIP aims to analyze students' practice behavior and its impacts on students' performance and retention. Studies have shown that memories weaken over time, and the most significant drop in retention happens just a day after you learn something new. Ebbinghaus's forgetting curve suggests that people tend to forget newly learned materials in days or weeks when there is no attempt to review the learned materials. To prevent forgetting, we need to reinforce what we learn by using techniques such as spaced retrieval practice and interleaving. We did that in an Introductory Computer Science Course (CS1). We used quizzes as low-stakes retrieval practice activities (RPAs) that students could take as many times as they want during the semester, along with the U-Behavior teaching and learning method. The U-Behavior method is composed of an application that generates personalized visualizations of student's study behaviors and self-reflections to help students improve their study behaviors to reinforce long-term learning. In a previous study, with data from a Fall 2021 CS1 course, we found a statistically significant increase in final exam scores, final coding exam, and final course grades for students who practice desirable behaviors, spacing and mixing their practice over the semester, compared with students who did not. This study presents a work in progress that uses data from Fall 2023 CS1 course to analyze if the increase in grades persists in a different cohort of students for the same course as well as to extend the previous study by monitoring students practice behaviors in a subsequent class, CS2, that is being taught this Spring 2024 semester. Of 94 students registered in CS1 Fall 2023, a total of 56 consented to participate in the study. Of the 56 students, 18 demonstrated desirable behaviors by spacing and interleaving their practice during the semester, and 38 did not. Initial analysis of Fall 2023 CS1 data showed a statistically significant increase on final exam and final course grades for students who practice desirable study behaviors. Contrary to Fall 2021 CS1 data, no statistical significance was found regarding final coding exam grades for Fall 2023 CS1 students between students who practice desirable behaviors and those who didn't. We analyzed if students who demonstrated desirable practice behaviors are continuing practice in a spaced and interleaved in the subsequent class, CS2. Initial findings and future directions are presented.

Students in entry level CS courses come from diverse backgrounds and are learning study and time management skills. Our belief for their success is that they must master a growth mindset and that the final grade should represent their final mastery of topics in the course. Traditional grading systems tend to be too restrictive and hinder a growth mindset. They require strict deadlines that fail to easily account for student accommodations and learning differences. Furthermore, they run into averaging and scaling issues with 59% of a score counting as failing, making it difficult for students to redeem grades even if they later demonstrate mastery of topics.