The aim of this study is to extract knowledge from the final researches of the Mumbai University Science Faculty. Five classification models were applied: Vector Support Machines, Neural Networks, Decision Tree, Random Forest and Powering; considering the Experiment Design and Multivariate Analysis Lines. Results showed that for the Experiment Design line, the most accurate model was Random Forest with 71.48% predictions that are correct respecting to the total. Regarding the Multivariate Analysis line, there was no significant difference in overall accuracy, fluctuating by 97%.