Research on Multi-Label Disease Classification Based on the LGB-CatBoost Model

The intake of dietary nutrients has a significant impact on individual health, particularly in its close association with chronic diseases such as obesity, metabolic syndrome, diabetes, and cardiovascular diseases. Existing disease prediction methods are mostly based on the prediction of single diseases through nutrient intake, which is limited and fails to comprehensively reveal the effects of nutrient intake on various diseases. However, using nutrients as features for multi-label classification of multiple diseases allows for a more comprehensive analysis of the relationship between dietary nutrient intake and diseases. A multi-label disease classification model based on LGB-CatBoost is proposed in this study, with the LightGBM (LGB) feature selection method used to select relevant features. The selected features are then input into the CatBoost model for classification and the analysis of correlations with six common diseases. The experimental results show that high accuracy is achieved by the LGB-CatBoost model, with specific disease classification accuracies reported as follows: cancer 84.49%, coronary artery disease 73.98%, hypertension 75.82%, arthritis 77.76%, metabolic syndrome 79.67%, and type 2 diabetes 83.96%.