Multidimensional analysis based on historical Olympic data-medal prediction and trend exploration

The analysis of historical Olympic data holds significant importance for research and practical application. This study constructs an Olympic athlete dataset using Kaggle and a self-built dataset. Employing data cleaning and multidimensional data mining, it investigates the relationship between height, weight, age, and medal acquisition. The RoBERTa-CNN-BiLSTM-ATT-CRF model is introduced to effectively extract and categorize entities like athlete and event information, enhancing analytical accuracy. This model shows improvements in P-value, R-value, and F1-value compared to previous entity recognition models, particularly in F1-value. The results indicate the model's capacity to capture the interplay of factors such as athletes' height, weight, and age with medal outcomes. These findings reveal the impact of scientific training and physical conditioning on competitive performance. This study emphasizes the role of physical well-being in sports, aligning with the Olympic principles of 'green, health, and sustainable development. By summarizing factors influencing winning, training a predictive model, and offering medal optimization recommendations, it provides data-driven support for scientific training and holistic health. The sophisticated RoBERTa-CNN-BiLSTM-ATT-CRF model advances the state-of-the-art in entity recognition through its innovative fusion of multi-level semantic representation and attention-driven contextualization, ultimately contributing to a more informed and sustainable approach to the Olympic Games.