As a team sport, basketball has the characteristics of strong antagonism, appreciation and skill. In basketball, players usually have to face the huge pressure from many parts of the body, especially in the attack, defense, sudden action. In particular, the athletes need to be very professional in taking off, defending, breaking through and breaking into the basket. At the same time, because basketball has the characteristics of fast, quiet and accurate, the probability of upper limb movement disorder after injury is relatively high. Basketball is widely loved by young people today. With the progress of society, basketball has entered the life of the public, but there is also spinal cord injury. Spinal cord injury is a serious disease that cannot be cured at present, with a high disability rate and relatively serious consequences. For spinal cord injury, regular exercise rehabilitation training is a very effective treatment. Therefore, this paper proposed to use machine learning algorithm to predict spinal cord injury in basketball sports and analyzed the effect of rehabilitation treatment of upper limb motor disorders. The research shows that under the same other conditions, when FMS (Functional Movement Screen) ≥ 14 points, the injury rate of athletes is higher, and when FMS ≤ 14 points, the injury rate of athletes is lower. It shows that machine learning can effectively predict spinal cord injury in basketball.