Optimizing Neural Network Hyperparameters Using Genetic Algorithms for Predicting Student Adaptability in Online Education

Abstract

Predicting student adaption is a crucial component of studying online learning material. Machine learning algorithms are crucial in this situation. Deep learning is a fundamental concept in machine learning algorithms. This work used Python in the Jupyter Notebook environment to implement the deep learning approach for forecasting students' adaptation to online learning. The Keras and Tensorflow libraries were used to construct a neural network model using the Kaggle dataset. The data is divided into testing data and training sets and utilize the Keras plot_model utility method to visualize the neural network model. Construct the deep learning model with two hidden layers, each employing randomly picked activation functions from relu, sigmoid, tanh, elu, and selu. Additionally, include one output layer with the softmax activation function. After undergoing a fine-tuning procedure until the alterations stabilized, this model achieved an accuracy of 89.63%.