Liver disease is a major global health concern, requiring timely diagnosis and intervention to improve patient outcomes. In this study, we use clinical data to predict the stage of liver disease through a thorough data analysis. With the illness status acting as the target variable, the dataset comprises a number of demographic, biochemical, and clinical characteristics, including age, cholesterol, and copper. To find important patterns and connections in the data, we used exploratory data analysis, or EDA. To guarantee consistency, numerical features were normalized, missing data were imputed, and categorical information were encoded using one-hot encoding. We used oversampling approaches (SMOTE) to address the issue of class imbalance, which helped to balance the distribution of disease status. XGBoost (Extreme Gradient Boosting), a potent machine learning method renowned for its efficacy with structured data, served as the main classification model. Because XGBoost can handle missing values, manage imbalanced datasets, and provide feature importance rankings, it was selected. To maximize the model’s performance while preventing overfitting, we used GridSearchCV to do hyperparameter tuning, optimizing important parameters like n_estimators, max_depth, learning_rate, and subsample. The model’s capacity to predict the status of liver disease was thoroughly assessed using a variety of measures, including accuracy, precision, recall, F1-score, and ROC AUC. The finished model performed well, highlighting the significance of a strong, data-driven strategy for forecasting the stages of liver disease. This model emphasizes the need of early detection and individualized treatment planning for patients, as well as its insights into the fundamental factors impacting the evolution of the disease.