PENCARIAN KERNEL TERBAIK SUPPORT VECTOR REGRESSION PADA KASUS DATA KEMISKINAN DI INDONESIA DENGAN USER INTERFACE (GUI) MATLAB

Muhammad Ghazali(1*), Ramdani Purnamasari(2)


(1) 
(2) 
(*) Corresponding Author

Abstract


Poverty is a topic that is often discussed in various scientific study forums. Facts on the ground show that increasing development has not been able to reduce the increasing number of poor people. Statistical studies on poverty data need to be carried out to assist the government in mapping policy-making patterns. One of the variables in mapping poverty data patterns is the Poverty Depth Index. The poverty depth index is a measure of the average gap in the distribution of each population's expenditure on the poverty line. Many factors affect the poverty depth index, especially from health, human resources and economic indicators. Therefore, a statistical modeling is needed to analyze the factors that affect the poverty depth index in Indonesia. The poverty data used in this study were sourced from the 2019 SUSENAS data in the form of data with individual observations of all provinces in Indonesia. Several previous studies using the Support Vector Regression (SVR) method to estimate the Poverty Depth Index as a response variable with several variables from health and economic indicators showed a very good level of model accuracy. However, SVR is constrained by choosing the right kernel to find the optimum prediction accuracy. So it is necessary to create a user interface that automatically selects the best type of kernel to facilitate the modeling process. The user interface will also help users to use the SVR even if they do not know the programming language. This study aims to: (1) produce a statistical analysis that makes it easier to map the pattern of factors that influence poverty in Indonesia, (2) produce a user interface that makes it easier for users to analyze poverty data in Indonesia. The conclusion obtained from this study is that the most accurate estimation is to use a degree 1 Gaussian kernel (RBF) SVR model while using the Polynomial kernel is not enough to provide a good estimate.

Full Text:

PDF

Article Metrics

Abstract view : 16 times
PDF - 0 times

DOI: https://doi.org/10.26714/jsunimus.9.1.2021.1-8

Refbacks

  • There are currently no refbacks.


Copyright (c) 2021 Jurnal Statistika Universitas Muhammadiyah Semarang

Editorial Office:
Department of Statistics
Faculty Of Mathematics And Natural Sciences 
Universitas Muhammadiyah Semarang

Jl. Kedungmundu No. 18 Semarang Indonesia

Published by: 
Department of Statistics Universitas Muhammadiyah Semarang

View My Stats

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.