Scholarlink

<p>Predicting the market value of professional soccer players is a central challenge in sports analytics because player valuations follow nonlinear patterns and fluctuate with market conditions. This study presents a hybrid predictive framework that integrates Extra Trees Regression (ETR), Light Gradient Boosting Regression (LGBR), and Extreme Gradient Boosting Regression (XGBR) with two nature-inspired optimization algorithms, Black Widow Optimization (BWO) and the Chimp Optimization Algorithm (CHOA), which were selected for their effective global search performance in hyperparameter tuning. Optimized variants of the baseline models, including ETBW, ETCA, LGBW, and LGCA, were developed to improve predictive accuracy. Feature selection was applied during preprocessing to reduce multicollinearity and strengthen model reliability. Model performance was evaluated using RMSE, R², n10-index, SI, and U95. Among the models, the LGBR–CHOA hybrid (LGCA) achieved the best result with RMSE of 2.49&#xa0;million euros and R² of 0.981, indicating strong generalization and reduced uncertainty. These results promote the integration of metaheuristic optimization with regression-based valuation methods and thereby facilitate the creation of dependable and clear instruments for analyzing the football market.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Soccer players’ market value assessment via a multi-optimal framework analyzing the importance of deep features importance

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Jing Yang,
  • Bin Wei,
  • Yunjian Xia,
  • Hongjun Ma,
  • Feng Wang

摘要

Predicting the market value of professional soccer players is a central challenge in sports analytics because player valuations follow nonlinear patterns and fluctuate with market conditions. This study presents a hybrid predictive framework that integrates Extra Trees Regression (ETR), Light Gradient Boosting Regression (LGBR), and Extreme Gradient Boosting Regression (XGBR) with two nature-inspired optimization algorithms, Black Widow Optimization (BWO) and the Chimp Optimization Algorithm (CHOA), which were selected for their effective global search performance in hyperparameter tuning. Optimized variants of the baseline models, including ETBW, ETCA, LGBW, and LGCA, were developed to improve predictive accuracy. Feature selection was applied during preprocessing to reduce multicollinearity and strengthen model reliability. Model performance was evaluated using RMSE, R², n10-index, SI, and U95. Among the models, the LGBR–CHOA hybrid (LGCA) achieved the best result with RMSE of 2.49 million euros and R² of 0.981, indicating strong generalization and reduced uncertainty. These results promote the integration of metaheuristic optimization with regression-based valuation methods and thereby facilitate the creation of dependable and clear instruments for analyzing the football market.