The late discovery of breast cancer is the main reason for the low survival rate, so its early detection can lessen the risk and prevent progression, thus reducing mortality rates. Machine learning classification approaches play a significant role in predicting breast cancer in its early stages, which support the process of taking the appropriate treatment that leads to enhanced survival rates. However, selecting an adequate classifier, dataset and features are the keys to successfully improving prediction accuracy and performance. This paper has introduced an efficient ensemble classification model that is based on a stacking technique with feature selection method. The base classifiers of the proposed ensemble model are SVM, neural network, random forest, gradient boosting and KNN, while logistic regression is used as a meta classifier. An enhanced version of WBCD has been employed. WBCD is one of the most common datasets for breast cancer detection. A feature selection method has been applied and, accordingly, only 23 of the 30 features are selected. Only the features with a high influence on the prediction process are considered. Our proposed model shows a high accuracy rate that reached 98.4%. It outperforms the single classifiers and causes neglected classification errors. The proposed model has also been compared to some similar existing approaches and showed better performance.