Background: Routine sentinel lymph node biopsy in older breast cancer patients with favorable tumor biology is not recommended. However, cases must be evaluated on an individual basis to avoid under or over-treatment. Many nomograms have been developed to calculate the risk of nodal positivity, but machine learning (ML) is a novel tool that may improve the accuracy of nodal prediction. In this study, we developed a support vector machine (SVM) model to delineate factors indicative of sentinel lymph node positivity and refine individualized nodal risk assessment for this heterogeneous patient population. Methods: We conducted a single-institution comprehensive retrospective review of patients 70 years or older diagnosed with unilateral stage I–III primary breast cancer from January 2005 to January 2016. Patient data was partitioned into training and testing sets. A SVM model was developed to predict lymph node status using patients’ demographics, tumor stage, genetic profile, and imaging data. Primary outcome was model performance determined by area under the curve (AUC). Secondary outcomes were accuracy, sensitivity and specificity. Permutation feature importance (PFI) analysis and accumulated local effect (ALE) plots were used to evaluate significant predictors identified by the SVM. Results: We identified 1706 consecutive patients who met the study criteria with a mean age of 76±4.5 years. The plurality of patients were Caucasian (82%), had ER+ (86%), PR+ (70%), HER2- (87%) stage I (72%) breast cancer. Sixteen percent of patients (n = 271) had a positive sentinel lymph node biopsy. The SVM model demonstrated good discriminatory performance for predicting sentinel lymph node positivity with mean AUC of 0.70 (95%CI, 0.62-0.77), mean accuracy of 84% (95%CI, 80-88%), mean sensitivity of 61% (95%CI, 57-66%), and mean specificity of 62% (95%CI, 52-73%). PFI and ALE identified higher disease stage, younger age, family history of breast cancer, margin status, estrogen and progesterone receptor positivity as independently associated with high risk of sentinel lymph node positivity. Conclusions: The proposed ML model accurately identified sentinel lymph node status in older patients with breast cancer. This model holds promise for counselling patients as to the potential risk for node positive disease which may impact surgical and adjuvant therapy recommendations.