Performance of student’s efficiency is a serious problem that happens usually in colleges. There are many techniques to predict their performance. To apply necessary steps for the weak students, the predictions have been helpful for selecting weak students who will be less than average and who can assist the management, so that they can perform well in their next tests. The selection of the best classification algorithm for a given dataset is a very widespread problem and also a complex one. In this work we focus on the accuracy to assess the classification performance. First we compare the accuracy of different classification algorithms used in data mining on multiple datasets and select an algorithm which has high accuracy for all classes to ensemble them with the help of a method by name, weak classifier to get a combination of multiple classifier. Then we measure the accuracy of combination of weak classifiers (Naïve Bayes, OneR and Decision Stump).We present results which display that the combination of multiple classifiers using vote method is more accurate than other single classifiers. There are other techniques also which are used in preprocessing of dataset like re-sampling to produce more accurate results. The results with the re-sampled data set proved to be more efficient and accurate compared to the original dataset. © Research India Publications.