Speed-reading courses are designed to improve their students’ reading speed and comprehension. The use of e-learning environments enables data collection that helps in the assessment of students and predictive analyses. Student dropout prediction is among the popular problems in this context. This study presents a neural network-based prediction model that identifies impending dropouts from a speed-reading course. Assessment scores obtained in the last ten sections were analyzed to predict dropouts who will not proceed to the next level. Despite the challenge of predicting short-term dropouts, the tests resulted in an accuracy of 78.24% on average. Moreover, 56.58% of predicted students dropped out before the next level, while 52.67% of students were successfully identified just before the dropout.