Proteins execute and coordinate cellular functions by interacting with other biomolecules that help in life sustenance. In order to use the fundamental knowledge of protein functions for industrial or medical applications such as in protein engineering and drug design, it is essential that they be understood in detail. Identification and characterization of functional sites is an important step towards this end. These sites often comprise of residues that bind with ligands, participate in catalysis etc. However, because the occurrence of these sites is very few as compared to their counterpart, their accurate identification has been challenging. Knowledge guided intuitive approaches such as machine learning could aid rapid and robust identification of protein functional sites. It can be done, provided the nature of imbalance in data is addressed adequately. Machine learning approaches have gained popularity in this context for developing ab initio methods of protein functional site prediction. Here biological properties are encoded appropriately in a computational architecture for predictions. In this project biological data analysis and processing leading to the development of predictive models for prediction of protein functional sites using machine learning is proposed.