An Efficient High Dimensional Gene Expression Based Multi-Class Cancer Prediction Model using Ensemble Learning

Abstract

Cancer continues to be a predominant worldwide health concern, and early, non-invasive identification significantly improves survival rates and treatment results. Conventional diagnostic techniques, including tissue biopsies and imaging, are often invasive, costly, and time-consuming. Recent improvements in next-generation sequencing facilitate the examination of gene expression data at the molecular level. Nevertheless, this data is high-dimensional and noisy, making manual analysis impossible. This paper presents a machine learning framework for multi-class cancer categorization using Tumor-Educated Platelet (TEP) gene expression data. The dataset undergoes preprocessing using normalization, feature selection, and Principal Component Analysis (PCA) to reduce data dimensionality. Several supervised models, including Support Vector Machine (SVM), Random Forest, Logistic Regression, and XGBoost, were trained and refined by cross-validation. The performance is assessed using accuracy, precision, recall, F1-score, and ROC-AUC metrics. Experimental findings demonstrate that ensemble models, notably Random Forest and XGBoost, surpass other classifiers in managing high-dimensional biological data. The finished model is then deployed via a web-based application for real-time, user-friendly cancer prediction. This study presents a scalable and noninvasive method for precise cancer categorization that facilitates future clinical decision-making systems.

Keywords

Article Information