In this paper, a novel hybrid approach was presented combining random forest (RF) and k-nearest neighbors (kNN) for the classification of breast cancer data. RF is selected for its robustness against overfitting and its ability to handle high-dimensional data effectively, providing a measure of feature importance and generalizing well due to its ensemble nature. kNN is chosen for its simplicity and effectiveness in capturing local data patterns. Our hybrid RF-kNN approach involves feature importance weighting in kNN, dynamic k selection, polynomial feature expansion, and ensemble output combination. The Wisconsin breast cancer database (WBCD) is used for experimentation, evaluated using 10-fold cross-validation. Performance metrics include accuracy, precision, recall, and F1-score. The results demonstrate that the hybrid RF-kNN model outperforms individual models, achieving superior performance across all metrics and data splits. This highlights the robustness and effectiveness of the hybrid model in reducing false positives and correctly identifying patients with breast cancer, making it a reliable model for breast cancer detection.