ABSTRACT: These technologies, which are frequently employed in computer classrooms, are referred to as Online Judge (OJ). They are capable of impartially and expeditiously evaluating pupil work. Typically, this evaluation system generates a single outcome when a rubric is employed to determine whether a submission satisfies the task's requirements. Professors and students would both benefit from a greater degree of control over the overall evaluation of the project, as the automatic assessment system may fail to recognize certain aspects of exceptional academic achievement. We will utilize OJ data to provide instructors and children with real-time feedback to assist them in overcoming this obstacle. Multi-Instance Learning and essential machine learning methods, which are learning-based techniques that replicate student behavior, may generate more precise assessments. The model supports the hypothesis by accurately predicting a student's outcome, which is either passing or failing an assignment, based solely on the patterns of behavior shown in OJ entries. Teachers and students equally can benefit from this method, as it simplifies the presentation of more pertinent information, including student profiles and at-risk groups.