Enhancing Cancer Prediction Accuracy: Investigating Novel Machine Learning Approaches for Early Detection

Abstract

The ongoing impact of cancer on global health emphasizes the critical importance of advancing predictive modelling capabilities to enhance early detection, diagnosis, and treatment outcomes. This study develops a synthetic dataset encompassing clinical features for lung, breast, and prostate cancers, comprising 3,000 samples with diverse attributes, and investigates the performance of three advanced machine learning algorithms. Our evaluation, using Accuracy, Precision, Recall, F1 Score, and AUC-ROC metrics, showed Transformer Models leading the pack with 92% accuracy and 0.96 AUC-ROC, underscoring their ability to uncover complex relationships within the data. With accuracy and AUC-ROC scores of 89% and 0.91, respectively, AutoML frameworks demonstrated their potential as a user-friendly and adaptable solution for machine learning tasks. Our study showcases the promise of machine learning in predicting multiple cancer types, emphasizing the importance of algorithmic innovation in enhancing predictive power. We advocate for integrating advanced models, such as Transformers and GNN s, into real-world clinical applications, and highlight the utility of AutoML for streamlined model development.

Ähnliche Arbeiten