Neural Network Optimization for Prediction of Student Study Period

Abstract

The student's study period's in a university was significant in implementing higher education goals and study programs to improve accreditation level. The student's study period's prediction can make higher education institutions' foundation in making future policies. Several factors in implementing students during their studies, including the cumulative achievement index (GPA), affect the study period. Furthermore, the institution often does not consider the conditions or the student's study period's predictive value at its campus. A neural network (NN) is a prediction or classification method that previous researchers have widely used because it is reliable in solving prediction problems. The main problem with improving the accuracy of the NN is the tuning parameter. The neural network model has algorithms for optimization, namely, Particle Swarm Optimization (PSO) and Genetic Algorithm(GA). Based on the experiments and analyses that have been done, the accuracy has been obtained in the GA (GA-ANN) Neural network model with an accuracy score of 71.4%. The score is gained from the parameter specification number of epoch 5, mutation rate = 0.9, layer size 20, tanh activation function, adam solver, and 1000 maximum iteration.