Gene Expression-Based Cancer Classification using Machine Learning

Breast cancer is still one of the primary causes of mortality among women, making it critical for timely diagnosis and improved treatment. Molecular pattern analysis via gene expression provides an excellent opportunity for discovering patterns that correspond to cancer development; nevertheless, high dimensionality poses some difficulties for classification. In order to cope with the issue, the current study suggests implementing a machine learning-driven model for breast cancer classification based on BC-TCGA dataset. Preprocessing was conducted through missing value imputation with Imperative SVD using normalization. Feature selection was done by means of an AGA-based algorithm combined with MIM. As a result, various classifiers such as SVMs, RF, Logistic Regression, and XGBoost were trained using a selected set of genes. The suggested pipeline produced accuracy above 90. This work demonstrates that implementation of both feature selection and machine learning contributes considerably to breast cancer prediction with gene expression.