Diabetes mellitus (DM) is a chronic metabolic disease that is typically analyzed through various methods such as blood glucose tests, oral glucose tolerance tests, and HbA1c measurements. Early detection is crucial for proper management and prevention of complications associated with this disorder, including kidney problems, cardiovascular diseases, and nerve damage. Machine learning (ML) and feature selection (FS) play a significant role in DM detection, employing data-driven techniques to leverage advanced models for identifying relevant features from datasets, including medical history, clinical measurements, and patient demographics. This paper presents the DM recognition using Marine Predators Algorithm with machine learning (DMR-MPAML) model. The DMR-MPAML model is developed to recognize DM by employing both FS and parameter tuning methodologies. Initially, the data undergoes Min-Max normalization to ensure uniformity. The model utilizes the Marine Predators Algorithm (MPA) for efficient feature selection, while feed forward neural network (FFNN) serves for DM detection. The parameter tuning approach, particle swarm optimization (PSO), enhances DM recognition in FFNN. DMR-MPAML model improves detection rates, streamlines recognition, and aids in risk assessment, leading to better patient outcomes. Validation on the PIMA Indians Diabetes dataset confirms its superior performance in DM detection.