Lymphoma is a malignant tumor that originates from the lymphatic system. This study aims to develop a risk prediction model for lymphoma using lymphoma cases extracted from the 2023 National Health Insurance Service (NHIS) database. The χ² test was used to examine differences in seven variables between groups. Subsequently, hyperparameter combinations were randomly tested for optimization. Four machine learning models were constructed after obtaining the optimal hyperparameters. Receiver operating characteristic (ROC) analysis was employed to evaluate the four machine learning models and select the most effective model. The relative importance of the seven variables was ranked using the machine learning algorithm. The data from the baseline characteristic table, comprising 3,670 participants, revealed that six variables differed significantly (p < 0.05). After hyperparameter tuning, the extreme gradient boosting (XGBoost) model achieved the optimal parameter combination, with a receiver operating characteristic value of 0.8567. The XGBoost model also demonstrated the highest area under the curve value (0.8161), validating it as the best-performing model, with a sensitivity of 0.9916, a specificity of 0.1188, a precision of 0.9086, an F1 score of 0.9483, and an accuracy of 0.9028. Finally, the machine learning model found that obesity (body mass index ≥ 30 kg/m²) exerted the greatest predictive contribution to lymphoma, with a relative importance score of 100. This study identified seven variables associated with lymphoma occurrence and developed a risk prediction model for lymphoma, providing valuable insights into the treatment of this disease.