A Machine Learning-Based Model for Comprehensive Assessment and Classification of Cross-Country Skiing Athletes

This paper presents a novel machine learning-based model for evaluating and categorizing cross-country skiers into five distinct levels. The model integrates Multi-Layer Perceptrons (MLP), Attention mechanisms, and Support Vector Machines (SVM) to comprehensively assess athletes based on three key categories of evaluation metrics: physical condition, psychological status, and professional skills assessment indicators. Each category of metrics is processed by separate MLP models to extract high-level feature representations. The extracted features from the MLP models are dynamically weighted and fused using an Attention mechanism, creating a comprehensive feature vector that reflects the overall potential of the athlete. This feature vector is then used as input to an SVM model for final categorization into one of the five skill levels. We employ a dataset of 1,000 athletes from professional training centers across Europe, North America, and East Asia. These athletes, balanced by gender(50% male, 50% female), age(18-35 years), and experience(1-15 years), 19 provide a diverse basis for model evaluation. Experimental results show that our model achieves an F1 score of 0.9306, representing an absolute improvement of over 0.10 compared to the best baseline method. The dynamic feature weighting and fusion capabilities of the Attention mechanism enable a more nuanced and precise aggregation of information from diverse metric categories, leading to more accurate and reliable athlete classification. Meanwhile, we have also proven through experiments that our model is superior to manual grading and more scientific.