Analysis of Track and Field Training and Performance Improvement Based on Big Data

Big data technology provides a new optimization path for track and field training. A training analysis system based on multi-source heterogeneous data was developed for sprinters in a certain provincial team. Deep learning algorithms and optimization models were applied for technical analysis and training intervention. Training load was optimized through deep reinforcement learning, and HRNet and ST-GCN networks were improved to achieve technical action analysis. A multi-channel feedback and Bayesian network decision-making system was constructed. Comparative analysis showed that big data-driven training improved athletes' 100-meter performance by an average of 0.105 seconds, reduced starting reaction time by 7.8%, improved step frequency stability by 14.2%, and reduced training related injury rates by 21.4% and training efficiency by 12.5% within a three-month period. Research has shown that personalized training programs can effectively shorten the training bottleneck period, reducing it from 8.3 weeks to 5.7 weeks. This study provides methodological support and practical reference for the scientific and precise intervention of track and field training.