The increasing failure rates in software programming courses highlight the need for effective intervention strategies. This study develops a learning analytics model to predict at-risk students in an introductory software programming course at a university in Ecuador. Three machine learning-based predictive models were implemented during a 16-week course to identify at-risk students at key moments (the start of the course, week 5, and week 10). Data from demographic, academic, and participation factors were analyzed. The predictive models demonstrated improved accuracy over time, reaching over 80% accuracy by week 10. Results from early interventions based on the predictions show that factors such as early platform engagement and exam performance significantly influence students’ risk of failure. The study concludes that early identification of at-risk students, coupled with targeted interventions, can effectively reduce course failure rates. Future research will explore reducing prediction times and enhancing model accuracy through additional algorithms and variables.