基于Logistic回归模型的乳腺癌诊断特征筛选

Feature Selection for Breast Cancer Diagnosis Based on the Logistic Regression Model

摘要: 本文主要利用Logistic回归的数学建模方法对乳腺癌诊断问题进行研究，首先基于数据特征，对数据进行清洗、缺失值处理和标准化处理，采用点二列相关系数和基于Logistic回归的递归特征消除法(RFE)分别筛选出15个关键特征构建模型。通过对比两种方法所选特征构建的Logistic回归分类模型，发现基于RFE所选特征的模型准确率达到97.89%，优于基于点二列相关系数所选特征的模型准确率，表现出较高的预测性能。由此不仅验证了机器学习在乳腺癌预测中的有效性，还为未来模型优化和病因探索提供了重要参考，助力临床医生实现“三早”预防。

Abstract: This paper mainly studies the problem of breast cancer diagnosis by using the mathematical modeling method of Logistic regression. Firstly, based on the data characteristics, the data is cleaned, missing values are processed, and standardized. The point-biserial correlation coefficient and the recursive feature elimination method based on Logistic regression (RFE) are used to select 15 key features to build the model. By comparing the Logistic regression classification models constructed with the features selected by the two methods, it is found that the accuracy rate of the model based on the features selected by RFE reaches 97.89%, which is better than the accuracy rate of the model based on the features selected by the point-biserial correlation coefficient, showing high predictive performance. This not only verifies the effectiveness of machine learning in breast cancer prediction, but also provides an important reference for future model optimization and etiological exploration, helping clinicians achieve “three early” prevention.