Diabetes is the most common non-transmissible disease. Millions of people worldwide are affected every year, who need early diagnosis and treatment to reduce complications associated with this non-transmissible disease. The major causes of this disease are genetic, high cholesterol and BP, an unhealthy diet, and an inactive lifestyle. Early traditional diabetes prediction techniques can be performed using various health-related parameters, such as age, gender, and BMI. This kind of process depends on clinical test results and time for prediction also high. Recently, diagnosing diabetes has benefited significantly from machine learning and prediction based on medical images and clinical data. This study examines many machine learning algorithms, such as decision tree algorithm, Random Forest approach, logistic regression model, Support Vector Machine, as well as, lastly, the Gradient Boosting algorithm, which are used to do early diabetes prediction. The Indian diabetic dataset, called the PIMA Diabetes Dataset, comprises 768 patient records of 8 attributes, including age, BMI, insulin levels, and glucose concentration, utilised to predict and evaluate the efficacy of various machine learning techniques. The results of various algorithms are compared to evaluate their performance. The Recall, Precision and Accuracy for different machine learning methods are calculated to assess the accuracy of models and select the optimal model.