In 2021, the International Diabetes Foundation estimated that 537 million people suffered from diabetes and that number is expected to rise to over 783 million by the year 2045. Type 2 diabetes (T2D) makes up 90% of diabetes cases. Machine learning research has been conducted to predict T2D risk to curb individual onset through intervention. Many of these machine learning studies involved the exclusive use of individual models such as Random Forest, Logistic Regression, and Neural Networks. The purpose of this study is to highlight contemporary research in the application of machine learning techniques for T2D prediction and to propose an alternative ensemble-based approach which uses confidence weighted voting of different base model classifiers. The ensemble methodology used in this study generated an average AUC metric of .846 over five different iterations during testing compared to individual models which produced AUC metric values ranging from .716 to .845. The ensemble performed better than a Bayesian Network, Neural Network, and Logistical Regression model. It performed marginally better than a Support Vector Machine. This highlights that “confidence weighted voting” based ensemble approach holds promise in producing improved predictive performance for type 2 diabetes prediction. Researchers and practitioners alike may achieve better predictive performance by applying such methods and should therefore seek to incorporate confidence weighted based ensembles into their standard machine learning procedures for type 2 diabetes prediction. Future research can seek to explore different statistical methods of using the confidence levels of models which compromise an ensemble in making final predictions besides the voting mechanism used in this study.