Machine Learning based Prediction of Dropout Students from the Education University using SMOTE

In past decade, there have been several students who have dropped out from the educational institutions and it is increasing rapidly. This has become one of the challenging factors for the educational institution. The students are getting into the institution and embarking their learning with several expectations and dreams. The expectations of the students have not fulfilled due to various factors such as staff, management, parents, course chosen etc., that make them drop from their registered curriculum. However, this has become the main issue for all educational institutes wherein several researchers introduced the technique of data mining for analysing as well as predicting the student's dropout. Therefore, this paper focuses on early finding of dropout variables as an advance by dimensionality reduction using feature selection and extraction methods. In feature extraction, there may be an occurrence of imbalanced data that may affect the significance of Machine Learning (ML) techniques. Thus, Synthetic Minority Oversampling Technique (SMOTE) is subsequently added with Principal Compound Analysis (PCA) whereas the oversampling of imbalanced data is managed to balanced dataset. Moreover, the 1,243 student's data have been collected and analysed using proposed PCA-S MOTE to allow for a more accurate forecast in case of dropout. Accuracy performance related to PCA-SMOTE has been 97.6% that is evaluated through confusion matrix parameter and compared with existing ML to find out the exact students who are not satisfied with their fulfilment in the environment of education institute.