Diagnostic Analysis of Depression Based on XGBoost Modeling

Abstract

This study aims to respond to the demand for research on the genetic mechanism of depression and to utilize gene expression data to construct an efficient depression diagnosis system. In view of the high-dimensional nature of gene expression data, cross-platform differences and limited sample size, this study first eliminates platform differences through Z-Score standardization, screens public gene features, and handles tag coding and missing values to ensure the consistency and completeness of the data. On this basis, the XGBoost binary classification model with AUC as the main performance index is constructed. The model gives full play to its feature selection and high-dimensional data processing capabilities to effectively cope with the situation of large number of features and small sample size. Key genes were identified through feature importance analysis, and the consistency and robustness of the model were verified using sensitivity analysis and stability test. This study provides a scientific and effective solution for clinical diagnosis of depression and precision medicine research.