Large language model vs. traditional machine learning: Evaluating predictive models for early detection of tumor relapse

Timilsina, Mohan, Buosi, Samuele, Torrente, Maria, Provencio, Mariano, Cobo, Manuel, Rodrıguez Abreu, Delvys, López Castro, Rafael, Carcereny, Enric, Curry, Edward, Nováček, Vít. (2025). Large language model vs. traditional machine learning: Evaluating predictive models for early detection of tumor relapse. Expert Systems with Applications, 283, 127641. https://doi.org/10.1016/j.eswa.2025.127641

Abstract

n this study, we evaluate the effectiveness of foundational artificial intelligence (AI) models, particularly large language models (LLMs), in comparison to traditional machine learning methods for predicting tumor relapse in patients with non-small-cell lung cancer (NSCLC). With a high recurrence risk in NSCLC, early and accurate prediction is essential for improving patient outcomes and guiding treatment decisions. Our analysis utilizes a dataset of 1,348 patients, examining the performance of traditional machine learning models such as Random Forest, alongside cutting-edge LLMs like Mistral-7B, LLaMA-7B, Falcon-7B, and GPT-based models. While the Random Forest model slightly outperforms Mistral-7B in precision–recall for relapse prediction, the comparable results suggest that both approaches offer valuable insights for early relapse detection. This study underscores the potential of integrating classical machine learning with foundational AI models to enhance predictive accuracy in cancer prognosis, providing pathways for more personalized medical interventions.

Funder

Science Foundation Ireland
European Commission

Publisher

Elsevier

Rights

Attribution 4.0 International

cb

Collections

Data Science Institute (Scholarly Articles)