Digit identification is crucial for understanding the image and processing that follows it with the recognition of various patterns because a machine cannot successfully recognize the handwritten digits. All the characters are recognized by optical recognition of characters (OCRs), which digitizes the printed documents and is a component of all the application of real-time. In the past, translating handwritten numerals to digital characters was a challenging problem. It is impossible to efficiently process the physical records without first converting them to digital counterparts, which is time-consuming and labor-intensive. Many algorithms and methods have been proposed as remedies for handwritten categorization over the years. With the help of support vector machines, convolutional network of neural, and the K-nearest method, this study tries to recognize solitary digits of handwritten. Following the implementation, training, and comparison of the models on the same dataset, it was shown that CNN, with a 98.59% accuracy rate, is the most efficient technique of machine learning which is used for categorizing all digits of handwritten.