Multilevel Logistic Regression and Neural Network-Genetic Algorithm for Modeling Internet Access

Wahyu Wibowo*, Shuzlina Abdul-Rahman, Nita Cahyani

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Logistic regression is one of the classical methods for classification. Meanwhile, neural network is the recent method for classification. Both methods are widely used in the supervised learning and competing to be the best methods in many classifications research. This paper aims to study the performance of both methods using data of youth internet access of East Java Province of Indonesia. The first method used is Multilevel Logistic Regression, a hierarchical model which is part of Generalized Linear Mixed Model (GLMM) where the response variable is influenced by fixed and random factors. The second one is Neural Network-Genetic Algorithm in which the weight optimization is performed by selecting the relevant input variables, the optimal number of hidden nodes, and the optimal connection weights. The result shows that Multilevel Logistic Regression produced a slightly better accuracy rate of 0.873 compared to Genetic Neural Network Algorithm with an accuracy rate of 0.871.

Original languageEnglish
Title of host publicationSoft Computing in Data Science - 5th International Conference, SCDS 2019, Proceedings
EditorsMichael W. Berry, Bee Wah Yap, Azlinah Mohamed, Mario Köppen
PublisherSpringer
Pages169-180
Number of pages12
ISBN (Print)9789811503986
DOIs
Publication statusPublished - 2019
Event5th International Conference on Soft Computing in Data Science, SCDS 2019 - Iizuka, Japan
Duration: 28 Aug 201929 Aug 2019

Publication series

NameCommunications in Computer and Information Science
Volume1100
ISSN (Print)1865-0929
ISSN (Electronic)1865-0937

Conference

Conference5th International Conference on Soft Computing in Data Science, SCDS 2019
Country/TerritoryJapan
CityIizuka
Period28/08/1929/08/19

Keywords

  • Accuracy
  • Multilevel Logistic Regression
  • Neural Network Genetic Algorithm

Fingerprint

Dive into the research topics of 'Multilevel Logistic Regression and Neural Network-Genetic Algorithm for Modeling Internet Access'. Together they form a unique fingerprint.

Cite this