Height prediction holds significant research value in the growth and development of children and adolescents, as well as in talent selection for athletes. The growth process in children is continuous and nonlinear, influenced by various factors such as bone age, chronological age, BMI, environmental, and genetic factors. Current research on height prediction methods primarily focuses on traditional approaches, such as those based on bone age or genetic height. With support from the Zhejiang Bone Age Research Center, this study constructed a dataset comprising 1,068 samples (615 males and 453 females). Each sample includes basic data such as gender, bone age testing time and value, date of birth, height, weight, adult height, and parents’ heights. Based on this dataset, this paper proposes a method for predicting adult height in children and adolescents using an ACPSO-SVR model. This model optimizes the hyperparameters of support vector regression (SVR) using an adaptive particle swarm optimization (PSO) algorithm, which adjusts the search strategy automatically based on the particle’s iterative position and count by optimizing inertia weight coefficients and the update method for learning factors. Experimental results show that the accuracy of the proposed algorithm in height prediction reached 92.15% for males and 91.11% for females. Compared to other height prediction models in recent studies, this algorithm exhibits higher prediction accuracy.