Predictive Model for Students' Academic Performance Using Classification and Feature Selection Techniques

At present, a lot of data related to students' academics are stored in digital form in the institution and if that data is processed properly to find some pattern related to student learning, then it helps to achieve a good result in future. So, data mining techniques help to find hidden information or pattern from student data. We have different data mining techniques to find the pattern and important information are classifications, clustering, regression and association rule mining algorithms. Here, in this paper, we implement different data mining techniques like naive Bayes, Random Forest, Decision Tree (J48), Multilayer Perceptron, Decision Table, JRip and Logistic Regression to predict the academic performance of the student. In any real world, the dataset contains a lot of features or attributes and all these attributes are not important or play any important roles in the mining process. So, the attributes in the dataset where the variance in the different values in the attributes are close to zero are deleted from the dataset, because it does not have any contribution to the mining process. We have different Feature Selection (FS) methods in data mining like Correlation Attribute Evaluator (CAE), Information Gain Attribute Evaluator (IGAE) and Gain Ratio Attribute Evaluator (GRAE). These features are used to remove those features which do not contribute to any mining process to predict the result. So, in this paper, we compare the prediction result of base classification algorithms with the classification algorithms used with some feature selection algorithm.