Abstract—Selection of the proper higher educational courses is absolutely necessary for the prospective students. Selecting appropriate courses are really cumbersome job for the students who are having less information about present trend of education relating to get placements or jobs and for better development in the future. In this paper, trend analysis and forecasting has proposed to predict the prospects of the selected higher educational courses in the field of computer science/technology. An online survey has done to get the dataset for analysis and there were altogether 151 data selected for the study. A Feed-Forward Artificial Neural Network model has proposed and the best network architecture has been selected among the top five NN considering the parameters like fitness value, AIC (Akaike’s Information Criterion) value, training, validation, test error values. The best network architecture is further analyzed using Levenberg-Marquardt (LM) and Conjugate Gradient Descent (CGD) algorithms for finding the accuracy of the trend. The study focuses on important input parameters during training of network architecture. Correct Classification Rate (CCR) for training and validation has been prepared to find the best network after a number of iterations. A comparative study<br> between the LM and CGD algorithm has primed with a focus on confusion matrix. This study recommends and predicts the future trends of the selected higher educational computer science/technology courses by using ANN.<br>
Authors: Dilip Roy Chowdhury, Deepanjan Sen
DOI: https://doi.org/10.5281/zenodo.5226838
Publish Year: 2017
