Diabetes is concern issue nowadays. It is rapidly growing disease with no cureness in which blood sugar increase in your body for longtime. There are many clinical methods are available for the detection of diabetes however they are time consuming and costlier. Therefor in lower income countries it becomes a limitation. As we discuss earlier like it is a time consuming and costlier many people remains undiagnostic which further causes savior complications and may also lead to fatalities. So it becomes essential to predict diabetes at early stage and prevent these complications and so in this research, we will work on a diabetes prediction by the use of different ML algorithms like LR, SVM, RF, NB, PCA. We use PID dataset for checking the model metrics measure. We used 8 parameters from PID dataset for model training and testing. Among all 4 algorithm and 1 technique, we found that SVM is outperformed in terms of all metrics measure