Volume 18, No. 6, 2021
Time Division Ensemble Learning Classifier For Efficient Student Performance Analysis Using Convolution Neural Network
Rasheed Mansoor Ali S , Dr.S.Perumal
Abstract
The problem of performance analysis and ranking of students has been well studied. There are number of methods available and suffer to achieve higher performance in various factors. To improve the performance, an efficient Time Division Multi Feature Ensemble Learning Classifier (TDMFEL) model is presented. The method maintains the history of students’ academic activity and performance achievements. The model monitors the academic activities and their performance in various corners. The history contains information related to each student which includes, number of seminars given, subject orient achievement, aptitude strength, extra curricular activities, knowledge on external world, leadership quality, presentation strength, number of admissions, number of placements and so on. Also, the model maintains the student records both on direct and online classes. Such traces are pre processed using Mandate Feature Analysis (MFA) to identify incomplete records. Further, the model extracts set of features like subject, score, learning activity, sports activity, seminar involvement, aptitude, achievements, external knowledge, number of presentations, number of appearances and so on. The features extracted are used in ensemble generation and trained with Convolution Neural Network. At the test phase, the neurons estimates support on Offline Learning Support (OLS), Online Learning Support (On LS), Offline Extra Curricular Support (OECS), Online Extra Curricular Support (On ECS), Sports Support (SS), Achievement Support (AS), Seminar and Presentation Support (SPS). According to the support measures computed, the models classify the student and predict the performance of the student. The proposed method improves the performance of prediction and ranking.
Pages: 3202-3219
Keywords: Performance Prediction, Student Performance Analysis, CNN, Ensemble Classifier, TDMFEL, MFA.