Objective Emerging evidence indicates that nearly 20% of patients with incompletely staged early-stage ovarian cancer develop to advanced stage because of lymph node metastases. The aim of the present study is to establish machine learning (ML) based predictive models for lymph node metastasis (LNM) in early-stage epithelial ovarian cancer (EOC). Methods The Surveillance, Epidemiology, and End Results (SEER) database was used to select patients diagnosed with early T classification epithelial ovarian cancer (EOC) between 2010 and 2015. The possibility of LNM was predicted by comparing the six ML algorithms. Model performance was compared in terms of accuracy, sensitivity, specificity, F1 score, and the area under the curve (AUC). The Shapley Additive Explanation (SHAP) analysis was employed to generate explanations and was presented as patient-specific visualizations. Results Screening of the SEER database yielded 3400 patients with early-stage epithelial ovarian cancer, and the data were divided randomly into a training set (70%), validation set (15%), and test set (15%). A grid search with 10-fold cross-validation was performed on the training set to tune the parameters. Overall, the Random Forest (RF) (accuracy of 79.8% and AUC of 0.878) was the best performing classifier and the Extreme Gradient Boosting (XGBoost) (accuracy of 77.2% and AUC of 0.857) demonstrated a similar high performance. The RF model performance was highly dependent on five top-rank features, including histology, grade, marital status, chemotherapy, and tumor size. SHAP analysis provided model-agnostic interpretation illustrating significant clinical contributors associated with risks of LNM in early-stage epithelial ovarian cancer (EOC). Conclusions The established ML-based prediction model for LNM in early-stage EOC is valid and valuable in improving clinical decision-making.