Today, Breast Cancer (BC) fatality risks are increasing dramatically. The detection of BC takes a long time due to the restricted nature of traditional systems. Despite the lack of treatments for breast cancer, the chances of survival are greatly influenced by early identification and diagnosis. A significant development and anticipated trend in the future of medicine is the use of machine learning (ML) for accurate medical diagnosis. Research on BC has greatly advanced our understanding of the disease over the past two decades, leading to safer and more effective treatment options. Despite this, numerous researchers have developed expert systems for BC early diagnosis. However, most expert systems usually fall short in successfully dealing with the class imbalance issue, as well as in performing systematic feature selection and efficient data pre-processing. To address these shortcomings, this study presents a “Machine Learning Based Expert System for Breast Cancer Prediction (MLESBCP)” for better BC prediction utilising ML analytics. The proposed system employs the KMeansSMOTE oversampling technique to balance the WBC dataset. The chi-square feature selection technique is utilised to identify the most essential feature of the WBC dataset to increase the model’s accuracy. The proposed model’s accuracy, F1-score, recall, and precision are compared to those of the previous single classifier models. The outcomes demonstrate that the MLESBCP achieved a maximum accuracy of 97.32%.