Scholarlink

<p>This study presents a dynamic machine learning framework for predicting the outcome of One Day International (ODI) cricket matches by analysing match progression at multiple game states. Each over is treated as a distinct match state, enabling real-time outcome prediction throughout the innings. Six key criteria are employed for classification, namely balls remaining, lead of Team A, wickets remaining, relative team strength, home advantage, and toss outcome. Feature extraction is performed using the League Championship Algorithm (LCA), which selects the most informative features from historical cricket data, followed by classification using a Back-Propagation Neural Network (BPNN). Experimental results demonstrate that the proposed model achieves an accuracy of 83%, a true positive rate of 0.81, a positive predictive value of 0.79, and an F1-score of 0.80 on the validation dataset, outperforming conventional prediction approaches by 5–10% across key performance metrics. The findings confirm the effectiveness of combining optimized feature extraction with neural network-based classification for accurate and interpretable cricket match outcome prediction.</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

Enhanced cricket match prediction using kernel methods for feature extraction and back-propagation neural networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月28日

查看全文
  • K. Dhinakaran,
  • S. Anbuchelian

摘要

This study presents a dynamic machine learning framework for predicting the outcome of One Day International (ODI) cricket matches by analysing match progression at multiple game states. Each over is treated as a distinct match state, enabling real-time outcome prediction throughout the innings. Six key criteria are employed for classification, namely balls remaining, lead of Team A, wickets remaining, relative team strength, home advantage, and toss outcome. Feature extraction is performed using the League Championship Algorithm (LCA), which selects the most informative features from historical cricket data, followed by classification using a Back-Propagation Neural Network (BPNN). Experimental results demonstrate that the proposed model achieves an accuracy of 83%, a true positive rate of 0.81, a positive predictive value of 0.79, and an F1-score of 0.80 on the validation dataset, outperforming conventional prediction approaches by 5–10% across key performance metrics. The findings confirm the effectiveness of combining optimized feature extraction with neural network-based classification for accurate and interpretable cricket match outcome prediction.