LIME-Explained Small-Scale Tabular Transformer Used for Improving the Classification Performance of Multi-Category Causes of Death in Colorectal Cancer

The timely identification and treatment of Colorectal cancer (CRC) are pivotal for enhancing patient prognosis. This paper presents a data-driven machine learning approach that aims to address the challenges in CRC prognosis. First an efficient feature selection method is used to address the problems posed by high and unbalanced data. A deep learning model based on the Transformer architecture is then used to process CRC small-scale tabular data. This model skillfully captures the inherent feature dependencies, leading to excellent performance in classification of causes of death in CRC patients. Additionally, the Local Interpretable Model-Agnostic Explanations (LIME) method is applied to interpret the model, providing explanations for both local and global level predictions. These elucidations promote our comprehension of the model's decision-making process and offer valuable insights for healthcare practitioners. The experimental results substantiate the validity and reliability of our methodology and demonstrate its potential extension to relevant prognostic diagnostic tasks and domains.