Education is necessary to improve an individuals’ life with quality and brilliance. The education system has changed from conventional Teaching/Learning to activity based learning. In spite of necessary measures for effective learning the education system is not able to reap the expected outcomes. Among various factors the most vital factor that determines the reputation of an Educational Institution is Students Performance. Although with wide literature available on performance of students, still it lacks necessary tools or approaches to address different challenges faced in identifying the low performing students for necessary pedagogical intervention and ensure successful completion of their graduation on-time. Predicting Student Performance much ahead is a challenging task and very less studies are available. Identifying probable risk students at an early stage is helpful for corrective measures by student, instructor and authorities. In this paper we present a novel Machine Learning approach and study the following: influence of student background on performance, Predicting First year Result, Predicting the performance of future semester from the progressive performance till date and Predicting the performance based on categorization of related Courses. The study is on real-time data of an Engineering College, we experimented with different predictive models to show that the proposed model achieves better performance with improved accuracy.