Penaksiran Parameter Model Regresi Logistik Multinomial Menggunakan Metode Optimasi Stochastic Gradient Descent dengan Quasihyperbolic Momentum pada Dataset Besar

Abstract

In multinomial logistic regression parameter estimation using maximum likelihood estimation, numerical approach with optimization method is used to get the best estimated parameters of the mentioned model. The optimization method of stochastic gradient descent (SGD) and the quasi-hyperbolic momentum (QHM) variant are going to be studied. Based on the dataset that has been studied, it has been found that the accuracy of classification using SGD with QHM is higher than using plain SGD.