Predicting Student Performance using the Application of Linear discriminant analysis

Understudy execution examination and proposing future strategy is need of great importance. To this end, this paper presents CLDA mechanism to classify the students into weak, fail and dropout students. Overall procedure of proposed system is portioned into phases. First phase consists of pre-processing mechanism in which infrequent values analysis and missing values handling. Predicting student performance should be fast and to achieve this objective, clustering mechanism is applied. Clustering is based on Euclidean distance. Values having lower distance are grouped into similar clusters. Looking into such clusters consumes least amount of time. Overlapping of clusters is tackled in the proposed mechanisms using CLDA(Clustering and linear discriminate analysis) and hence boundary value analysis is efficiently performed. ANN based training and classification is applied to predict student performance and suggesting comments for student future enhancement.
Techniques: Pre-processing by eliminating null values, clustering using clustering based CLDA, classification using ANN.
Parameters: Classification Accuracy, Specificity, sensitivity, F-Score
The mechanism employed reduces execution time and increase classification accuracy. In addition, boundary value analysis is efficient and hence distinguishment between pass, failed and withdrawn students.