Diabetes is becoming an increasingly serious global health and economic problem affecting 589 million people as of 2024 and 643 million people by 2030. Early and proactive measures are required to prevent the spread of this illness. The current study attempts to solve this issue by implementing an explainable, machine learning-based model for predicting diabetes, which used the Behavioral Risk Factor Surveillance System dataset, with more than 250,000 entries, supplied by Center for Disease Control and Prevention. In order to train the model, the dataset was prepared by applying feature engineering, normalization, and class imbalance issues have been addressed by means of applying random undersampling. A logistic regression model was built using the seven features that were chosen according to their mutual information values. These features were General Health, Blood Pressure, Age, Body Mass Index, Cholesterol, Difficulty in walking, and Income. As a result, the proposed model obtained 74.72% accuracy, 76.63% recall and ROC-AUC score of 0.75. It has demonstrated satisfactory performance and strong clinical applicability. Additionally, a user-oriented web interface was created for applying the aforementioned model to perform a fast and interpretable diabetes risk assessment with minimal input data. The current study improved upon previous works by including model explainability, high scalability and deployment due to creating the user-oriented web interface in the process.