Scholarlink

<p>Now, Breast Cancer is a serious health issue in the world, highlighting the need for clear and effective detection methods. Genetic algorithms (GAs) are employed to enhance the performance of machine learning models, including Random Forest (RF), Decision Tree (DT), Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and XGBoost, in this study. The primary aim is to improve feature identification, thereby boosting the accuracy and interpretability of these models. By using genetic algorithms, we can identify the most suitable features, allowing the models to focus on the most important data attributes, which in turn improves diagnostic accuracy. To ensure that the decisions made by the models are transparent and trustworthy, we implement Local Interpretable Model-agnostic Explanations (LIME), which help clarify the reasoning behind each model’s predictions. The research evaluates the accuracy of the models both before and after applying GA-based feature selection, revealing a notable improvement with the use of GA. For instance, RF and XG-Boost accuracies improved from 96.49% and 95.61% to 99.12% after GA-based feature selection. Furthermore, we assess the fidelity score of each model, which indicates how well the explanations generated by LIME align with the model’s actual predictions. Our results show that when models like RF and XG-Boost were improved with GA for feature selection, they showed high fidelity. This means that they can be used in clinical settings where both performance and readability are important. Even simpler models, such as DT and NB, when improved with GA, show competitive effectiveness with greater transparency, making them suitable for situations that require clear and understandable decisions. Unlike prior studies that used GA or LIME independently, this work presents a unified GA-driven feature optimization and LIME-based interpretability framework that simultaneously enhances predictive performance and clinical transparency.</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

Explainable breast cancer diagnosis: integrating genetic algorithms with LIME-based machine learning

  • 期刊论文

  • 发表日期：

    2025年12月26日

查看全文
  • Zulfikar Ali Ansari,
  • Md Shamsul Haque Ansari,
  • Ahmed Khan,
  • Hemlata Pant,
  • Sheikh Fahad,
  • P. Venkata Hari Prasad

摘要

Now, Breast Cancer is a serious health issue in the world, highlighting the need for clear and effective detection methods. Genetic algorithms (GAs) are employed to enhance the performance of machine learning models, including Random Forest (RF), Decision Tree (DT), Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and XGBoost, in this study. The primary aim is to improve feature identification, thereby boosting the accuracy and interpretability of these models. By using genetic algorithms, we can identify the most suitable features, allowing the models to focus on the most important data attributes, which in turn improves diagnostic accuracy. To ensure that the decisions made by the models are transparent and trustworthy, we implement Local Interpretable Model-agnostic Explanations (LIME), which help clarify the reasoning behind each model’s predictions. The research evaluates the accuracy of the models both before and after applying GA-based feature selection, revealing a notable improvement with the use of GA. For instance, RF and XG-Boost accuracies improved from 96.49% and 95.61% to 99.12% after GA-based feature selection. Furthermore, we assess the fidelity score of each model, which indicates how well the explanations generated by LIME align with the model’s actual predictions. Our results show that when models like RF and XG-Boost were improved with GA for feature selection, they showed high fidelity. This means that they can be used in clinical settings where both performance and readability are important. Even simpler models, such as DT and NB, when improved with GA, show competitive effectiveness with greater transparency, making them suitable for situations that require clear and understandable decisions. Unlike prior studies that used GA or LIME independently, this work presents a unified GA-driven feature optimization and LIME-based interpretability framework that simultaneously enhances predictive performance and clinical transparency.