Performance Prediction of Computer Science Students in Capstone Software Engineering Course Through Educational Data Mining

Educational data mining has been extensively used to predict students’ performance in university courses to plan improvements in teaching and learning processes, achieve academic goals, and timely support interventions. Computer Science (CS) courses focus on promoting problem solving skills through writing of software code and developing solutions using computing technologies. Within a four-year CS curriculum, the sequencing of courses is deliberately designed so that knowledge gained in a prerequisite lower level course is critical for success in upper-level courses. Overall, the CS curriculum prepares the students for a capstone experience in a final year Software Engineering (SE) course. The student success in SE course is dependent on skills such as requirement analysis, design, implementation, and testing gained in lower-level prerequisite courses. In this paper, we analyze grades data of 531 students in all under-graduate CS courses at a public university in the United States over a period of 8 years (2010 to 2018). Statistical analysis techniques including multiple linear regression, Pearson product-moment correlation coefficient, and paired samples t-test are used to analyze the data. The performance of students in SE course is investigated based on their grades in sequence of prerequisite courses including CS I, CS II, Data Structures and Object-oriented Programming. These prerequisite courses teach and test fundamental and advanced programming skills essential for success in SE course. The analysis shows CS II is a significant predictor of students’ success in the SE course. We also investigate the relationship between study of theoretical concepts and their application by examining the correlation between CS II (theory) and Data Structures (application) courses. Results shows a strong and positive correlation between students’ academic performance in the Data Structures course and CS I. We also observe the correlation between CS I and CS II. CS I builds fundamental concepts such as syntax, data types, control structures, selection statements, functions, and recursion while CS II focuses on advanced tools to use the concepts studied in CS I for problem solving. The results indicate a significant difference in mean grades in both courses. Conclusion, interpretations, and implications of these findings for the CS students will be discussed in detail in the full paper.