<p>This study examines the impact of discussions, presentations, and peer evaluations on predicting students' academic performance within a flipped classroom-based blended learning environment in university physics experiments. A total of 208 undergraduate students participated in this semester-long study, divided into an experimental group (<i>n = </i>122), which engaged in structured group discussions, presentations, and peer evaluations, and a control group (<i>n = </i>86), which followed a traditional blended learning approach. Analyses using independent sample t-tests, propensity score matching, and ANCOVA indicated that the experimental group achieved significantly higher academic performance. Correlation and hierarchical regression analyses identified six key predictors of academic success: activity performance, experiment reports, online quizzes, page views, initial scores, and participation in structured discussions and presentations. The analysis indicated that activity performance was the most crucial factor, highlighting the critical role of interactive and collaborative activities in promoting deeper learning and improved outcomes. The research also demonstrated the ability to predict students' academic achievement, and data that combined online and offline factors exhibited superior prediction accuracy. Additionally, starting in the fourth week of the course, the study demonstrated that it could predict students' final academic accomplishment and that prediction accuracy rose with course length. These findings underscore the value of integrating discussions, presentations, and peer evaluations within a flipped classroom-based blended learning framework to enhance student engagement, facilitate early academic performance predictions, and improve overall learning outcomes in university physics experiments.</p>