The prevalence of childhood obesity in China has recently become increasingly severe, and intervention measures are needed to stop its growth. Currently, there is a lack of assessment and prediction methods for childhood obesity. We develop a predictive model that uses currently measured predictors [gender, age, urban/rural, height and body mass index (BMI)] to quantify children's probabilities of belonging to one of four BMI category 5 years later and identify the high-risk group for possible intervention. A total of 88,980 students underwent a routine standard physical examination and were reexamined 5 years later to complete the study. The full model shows that boys, urban residence and height have positive effects and that age has a negative effect on transition to the overweight or obese category along with significant BMI effects. Our model correctly predicts BMI categories 5 years later for 70% of the students. From 2018 to 2023, the prevalence of obesity in rural boys and girls is expected to increase by 4% and 2%, respectively, while that in urban boys and girls is expected to remain unchanged. Predictive models help us assess the severity of childhood obesity and take targeted interventions and treatments to prevent it.