| 英文摘要 |
Purpose: This study applied the C5.0 decision tree algorithm to analyze women's 200m individual medley (IM) results from international competitions, including the World Aquatics Championships and the Olympic Games. The aim was to evaluate how split times in the four strokes (butterfly, backstroke, breaststroke, freestyle) predicted different ranking levels (G1: Finalists ranked 1st–3rd, G2: ranked 4th–8th, G3: ranked 9th–16th, G4: ranked 17th and beyond), and to build a practical model for training and talent identification. Methods: Data included 417 race performances from 2011 to 2024 (one record per swimmer per competition, using the last round swum). Stroke split times were input variables; the final ranking group was the target. The C5.0 algorithm was used to build the model, validated using cross-validation, and visualized to show decision logic and performance thresholds. Results: The model achieved an overall accuracy of 84.4%. Breaststroke was the most important predictor, appearing most frequently at key decision nodes, followed by freestyle and backstroke. Butterfly contributed mainly to secondary splits. Medalists typically swam breaststroke under 38.75s and freestyle under 30.86s. When freestyle was slower, better backstroke or breaststroke compensated. Non-qualifiers often showed weaker performance across multiple strokes. Conclusion: Breaststroke had the strongest influence on final rankings and should be a training focus. The C5.0 model offered clear visual guidance for coaches and athletes. Future work should expand variable types and apply the model to broader swimmer populations to improve scientific training and selection. |