Diabetes is one of the most chronic and deadly diseases. According to data from WHO in 2021, there were approximately 422 million adults living with diabetes worldwide, and this number is expected to continue to increase in the future due to various factors. Many studies have been conducted for early detection of diabetes by focusing on improving accuracy. However, a big problem in diabetes prediction is the selection of the right classification algorithm. This study aims to improve the accuracy of early detection of diabetes by implementing the Random Forest algorithm model. This research was conducted with the stages of data collection, data preprocessing, split data, modeling, and evaluation. This research uses the Pima Indian Diabetes data set. The results showed that the diabetes early detection model using the Random Forest algorithm produced an accuracy of 87%. This research shows that by using the Random Forest algorithm model, the performance of early detection of diabetes can be improved. However, there is still room for optimization of this performance, which is recommended for further research to carry out feature selection, data balancing, more complex model building, and exploring larger data.