Diabetes mellitus (DM) is a persistent sickness that may cause numerous difficulties. Machine learning methods are used to analyze and classification of diabetes. The learning-based calculations play an important role in supporting dynamic in infection conclusion and expectation. In this work, conventional categorization algorithms and artificial neural networks are researched for the diabetes dataset. Likewise, different execution strategies with various angles are assessed for the Naive Bayes, K-nearest neighbour, decision trees, Extremely Randomized tree, radial basis function and multi-layer perceptron (MLP) algorithms. It upholds the patient’s assessment that conceivably experiences the ill effects of diabetes later on. This paper gave that the feed-forward neural network algorithm multi-layer perceptron algorithm gives the most noteworthy expectation precision with the least Mean square error rate of 0.15. The multi-layer perceptron (MLP) gives the most reduced bogus negative rate and bogus positive rate with the most elevated region under the curve of 0.88.