Youth sports training is crucial for physical development, and Internet of Things (IoT) technology can realize scientific training, but it faces challenges such as high-dimensional data, noise interference, and unreasonable training intensity, so it is necessary to explore the data fusion model with higher adaptability to improve the training effect. This study proposes an adaptive bat algorithm optimized fuzzy clustering algorithm model for the characteristics of youth sports training data in the IoT environment. The model effectively avoids the problem that the traditional fuzzy C-mean clustering algorithm is prone to fall into local optimization by improving the velocity update formula in the bat algorithm and introducing the inertia weight coefficient adjustment mechanism based on the distribution entropy and average bit distance. Through the validation on the Iris and Wine datasets of UCI database, the results show that the clustering correct rate of this algorithm on the Iris dataset reaches 96.24%, which is 6.87% and 3.57% higher than that of the traditional FCM algorithm and GAKFCM algorithm, respectively, and that the correct rate on the Wine dataset reaches 94.76%, which is also better than that of the other two algorithms. Applying this algorithm to the analysis of 35,659 adolescents’ exercise behavior data, the algorithm successfully classified them into five class clusters and identified that 34.47% of the adolescents had regular exercise habits, 19.92% belonged to the category of frequent exercise but very seldom attendance, and only 10.02% hardly exercised. In the comparison of evaluation metrics, the proposed algorithm reaches 75.46%, which is significantly higher than 63.36% for K-Means and 67.23% for K-Means++. The study shows that the fuzzy clustering model optimized by the proposed adaptive bat algorithm can effectively deal with the complex data of youth sports training in the IoT environment, providing a reliable tool for data mining and personalized training program development.
