Scholarlink

Purpose <p>Breast cancer is one of the most common causes of mortality in women globally, and early detection markedly enhances treatment efficacy. This study conducts comparative analysis of several supervised machine learning algorithms for early detection of breast cancer, utilizing Wisconsin Breast Cancer Dataset (WBCD). The research question focuses on identifying the most efficient ML model for structured clinical datasets such as WBCD while maintaining high prediction accuracy.</p> Methods <p>This study employed a comparative analysis of six widely used supervised machine learning (ML) algorithms: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), and Multilayer Perceptron (MLP). The models were trained and tested using the WBCD from the UCI Machine Learning Repository. It is evaluated using essential performance indicators, e.g. accuracy, precision, recall, and F1-score. To ensure robustness and equity, hyperparameter optimization was performed using GridSearchCV, and three feature selection methodologies, such as correlation-based feature selection (CFS), sequential forward selection (SFS), and information gain (IG), were effectively employed to mitigate noise and redundancy.</p> Results <p>In all classifiers, SVM achieved the highest test precision at 98.24%, although MLP exhibited a competitive performance of 97.36%. Furthermore, interpretability was examined through the Explainable AI (XAI) approach, specifically Local Interpretable Model-Agnostic Explanations (LIME), to illustrate the impact of key features on model decisions, hence improving the clinical transparency of the predictive models.</p> Conclusions <p>Overall, the findings support the effectiveness of combining classical and deep learning methods with feature optimization and model interpretability for a reliable and explainable diagnosis of breast cancer. Future work aims to integrate domain-specific knowledge, such as biomarker data and histopathological features, to improve translational relevance.</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

Optimized and interpretable machine learning framework for early breast cancer detection

  • 期刊论文

  • 发表日期：

    2025年8月12日

查看全文
  • Zulfikar Ali Ansari,
  • Md Shamsul Haque Ansari,
  • Ahmed Khan,
  • NZ Jhanjhi,
  • Gopisetty Rathnamma

摘要

Purpose

Breast cancer is one of the most common causes of mortality in women globally, and early detection markedly enhances treatment efficacy. This study conducts comparative analysis of several supervised machine learning algorithms for early detection of breast cancer, utilizing Wisconsin Breast Cancer Dataset (WBCD). The research question focuses on identifying the most efficient ML model for structured clinical datasets such as WBCD while maintaining high prediction accuracy.

Methods

This study employed a comparative analysis of six widely used supervised machine learning (ML) algorithms: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), and Multilayer Perceptron (MLP). The models were trained and tested using the WBCD from the UCI Machine Learning Repository. It is evaluated using essential performance indicators, e.g. accuracy, precision, recall, and F1-score. To ensure robustness and equity, hyperparameter optimization was performed using GridSearchCV, and three feature selection methodologies, such as correlation-based feature selection (CFS), sequential forward selection (SFS), and information gain (IG), were effectively employed to mitigate noise and redundancy.

Results

In all classifiers, SVM achieved the highest test precision at 98.24%, although MLP exhibited a competitive performance of 97.36%. Furthermore, interpretability was examined through the Explainable AI (XAI) approach, specifically Local Interpretable Model-Agnostic Explanations (LIME), to illustrate the impact of key features on model decisions, hence improving the clinical transparency of the predictive models.

Conclusions

Overall, the findings support the effectiveness of combining classical and deep learning methods with feature optimization and model interpretability for a reliable and explainable diagnosis of breast cancer. Future work aims to integrate domain-specific knowledge, such as biomarker data and histopathological features, to improve translational relevance.