Construction of an intelligent recognition model for elderly fitness behavior preferences based on decision tree algorithms

Based on multi-source sensor data, this study constructs a computational model for identifying fitness behavior preferences among the elderly. Feature engineering enables structured representation of motion signals, while a hierarchical partitioning algorithm based on decision trees establishes nonlinear mappings between features and behavioral categories. The model undergoes training and cross-validation on standardized feature matrices, with classification performance evaluated via confusion matrices and ROC curves. Key decision pathways are revealed through feature contribution string diagrams. Experimental results demonstrate stable performance in recognition accuracy and interpretability, highlighting computational advantages for processing multidimensional behavioral data.