In medical decision making, the efficacy of classification, a vital component of decision support systems, can be significantly compromised by imbalanced class distributions within training data. This discrepancy is especially critical due to the heightened consequences of misclassification, particularly when a minority class is mistakenly classified as a majority class. Diabetes, posing a substantial public health challenge, underscores the urgency of addressing classification issues. The World Health Organization reports that 30% to 40% of diabetes cases remain undiagnosed, contributing to a 3% increase in diabetes-related mortality rates from 2000 to 2019. In 2019 alone, approximately 2 million deaths were attributed to diabetes and related kidney diseases. Consequently, effective intervention strategies are imperative. Managing imbalanced datasets in medical contexts presents significant obstacles to accurate classification, with profound implications for patient outcomes. To address this challenge, this research proposes leveraging machine learning and artificial intelligence techniques. The research comprises two main components: applying boosting and unbalanced bagging to a dataset containing diabetes-related information. Initial findings indicate that boosting yields promising results in classification accuracy. Building upon this success, we compare boosting with three additional technique Synthetic Minority Over-sampling Technique (SMOTE), Edited Nearest Neighbors (ENN), and Tomek links each sequentially applied to the dataset. Through rigorous experimentation and analysis, we aim to identify the most effective strategy for handling imbalanced datasets in diabetes classification. By enhancing our understanding of classification methods and their impact on imbalanced datasets, this research aims to contribute to improved decision-making processes in healthcare settings, ultimately leading to better outcomes for patients affected by diabetes.