Jurnal Eurekamatika
Vol 10, No 2 (2022): Jurnal Eurekamatika

Boosted Regression Trees

Wulan Dian Pramiesti (Universitas Pendidikan Indonesia)
Fitriani Agustina (Universitas Pendidikan Indonesia)
Rini Marwati (Universitas Pendidikan Indonesia)



Article Info

Publish Date
01 Dec 2022

Abstract

A regression tree is a data analysis technique that aims to determine the effect of all explanatory variables on the response variables. However, it turns out the regression tree method has a weakness that is the hierarchical structure allows for high variance predictors. The weakness of this regression tree method can be above using the Boosted Regression Trees (BRT) approach. The BRT approach uses a boosting technique to combine multiple regression trees in an additive manner so that the resulting variance is smaller than the previous variance. This study uses the BRT method for a case study of factors affecting the level of crime in East Java in 2018. Through 89 obtained optimal trees, the most important crime rate factors are the number of youth (model importance of 31.86%), followed by the percentage of poor population (11.14%), junior high school gross enrollment rate (9.67%), elementary school gross enrollment rate (8.98%), road stability (8.81 %), senior high school gross enrollment rate (8.67%), distance to the capital city of Surabaya (7.25%), open unemployment rate (7.02%), Gross Domestic Regional Product (GRDP) (5.32%), and population density (1.25%)..Keywords: Boosted Regression Trees, Boosting, Crime Rate, Regression Trees.AbstrakPohon regresi merupakan teknik analisis data yang bertujuan untuk mengetahui pengaruh semua peubah penjelas terhadap peubah responnya. Namun, ternyata metode pohon regresi memiliki kelemahan yaitu struktur hierarkisnya memungkinkan terjadinya varians yang tinggi. Kelemahan metode pohon regresi ini dapat diatasi dengan menggunakan pendekatan Boosted Regression Trees (BRT). Pendekatan BRT menggunakan teknik boosting untuk menggabungkan beberapa pohon regresi secara aditif sedemikian sehingga nilai varians yang dihasilkan lebih kecil dari nilai varians sebelumnya. Penelitian ini menggunakan metode BRT untuk studi kasus menentukan faktor-faktor yang memengaruhi tingkat kriminalitas di Jawa Timur tahun 2018. Melalui 89 buah pohon optimal yang terbentuk, diperoleh kesimpulan bahwa faktor yang memengaruhi tingkat kriminalitas tertinggi adalah jumlah pemuda (31,86%), diikuti oleh persentase penduduk miskin (11,14%), Angka Partisipasi Kasar (APK) SMP (9,67%), APK SD (8,98%), kemantapan jalan (8.81%), APK SMA (8,67%), jarak ke ibukota Surabaya (7,25%), tingkat pengangguran terbuka (7,02%), Produk Domestik Regional Bruto (5,32%), dan kepadatan penduduk (1,25%).

Copyrights © 2022






Journal Info

Abbrev

JEM

Publisher

Subject

Computer Science & IT Industrial & Manufacturing Engineering Mathematics

Description

Jurnal EurekaMatika (e-ISSN: 2528-4231, p-ISSN: 2776-480X) was first published annually on December 2013, and then since 2017 has been published twice a year, on May and November. JEM is a peer-reviewed Mathematics journal with its scope covers Algebra, Analysis, Statistics, and Applied Mathematics. ...