Deep Learning and Explainable AI Framework for Predicting Lung Cancer Severity

Abstract

Lung cancer is one of the main causes of cancer-related death worldwide, with high fatality rates due to late-stage detection, in many countries The situation is exacerbated by a lack of access to improved diagnostic technology and an increase in exposure to environmental risks. To address these problems, this study provides a machine learning-based framework for predicting lung cancer severity, with Explainable AI (XAI) techniques used to improve model interpretability- and clinical applicability. The study makes use of a large dataset, which includes demographic, lifestyle, environmental, and clinical variables. Following extensive data preprocessing, such as outlier removal and categorical encoding, the study creates and tests a variety of machine learning models, including XGBoost, Random Forest, Support Vector Machines, and Logistic Regression. XGBoost performed well, with an accuracy of 71.73% and an AUC-ROC of 0.7677. Explainable AI approaches, such as SHAP and LIME, were used to increase transparency and interpretability, guaranteeing that healthcare practitioners can trust and comprehend the model’s predictions. The data show that age, biomass fuel consumption, and tumor size had the most impact on prediction outcomes. The incorporate of XAI approaches not only increases the model’s transparency, but also provides doctors with interpretable information to help them make informed decisions, eventually improving patient outcomes. This work demonstrates the potential of using AI-driven predictive models to aid in early diagnosis and improve lung cancer management.