Predicting learning outcomes (LO) is an important problem in educational data mining (EDM). Despite extensive research, there remains a gap in determining the optimal timing for predicting early LO with acceptable accuracy. This study proposes a methodology for developing a predictive model using machine learning techniques to identify the appropriate time to forecast students’ course scores in a blended learning environment. The scope of the research includes students participating in the subject ‘Introduction to Computer Science’ at Hanoi National University of Education (HNUE). The experimental data includes data from the learning management system (LMS) and the face-to-face classroom. The research model is built based on machine learning algorithms, including Linear Regression (LN), Support Vector Regression (SVR), and Artificial Neural Networks (ANN). The research results show that the attributes most strongly influencing course scores are average practice score, midterm score, and major. The ANN model provides the highest efficiency with R2 = 0.6033. Early prediction of final exam scores can be made from week 7.