The invention discloses a learning scheme planning method based on machine learning, and the method comprises the steps: data collection, data analysis, learning scheme planning, execution condition feedback, learning by a student according to a planned task, and collection of the completion condition of the student, wherein the completion conditions of the students include the total task time consumption, the completion rate, the completion quality, the preference of task completion and the progress statistics of the academic achievements of all subjects. According to the scheme, the learningtask is formulated according to the detected current learning situation model and the learning scheme set in the database and pushed to the student, and the learning schemes formulated by different current learning situation models are different and more targeted; according to the completion condition of the student, the completion condition is fed back to a database for optimization, and then alearning scheme more suitable for the student is given; feedback and optimization are continuously and circularly carried out to form a virtuous cycle, the longer the same student uses the method, themore fitting the learning rule of the student is, the higher the efficiency is, and the higher the individuation degree is.