Hyperparameter determines the best learning curve on single
Published in Behavioural Sciences & Psychology
This primary study tries to explore the best machine learning model of hyperparameter tuning using single, multiple, and deep neural networks for predicting engineering student grade of Pokhara University Nepal in the year 2019. Generally, hyperparameter algorithms are adjusted from the data sets automatically when the model was designed. This is not a good idea for all sort of algorithm, with less flexibility of user choices. Therefore this article ties to meet the research gap between automatic calculation vs external model hyperparameter calculation on the same data sets applying comparison. The single neuron predicts 92 percent accuracy of the student's A, B, C, D grade of university examination with correlation with their internal marks respectively. The five-neuron multilayer neural network produces 72 percent accuracy and deep neural network with drop-out layer (32, 64) examines, the most suitable
I'm Yagyanath Rimal, a university teacher specializing in computer science. In addition to teaching C, C++, Java, and web applications at the bachelor's level, I also teach Research Methodology, Data Analytics, and Machine Learning to master's degree students. Currently, I am focusing on research in data analytics and machine learning. I've recently embarked on my journey in academia and am also interested in international collaborations and postdoctoral opportunities.