KLASIFIKASI ANGKA PENCURIAN DI RIAU DENGAN MULTIVARIATE ADAPTIVE REGRESSION SPLINES (MARS) DAN BOOTSTRAP AGGREGATING MARS
No Thumbnail Available
Date
2022-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Elfitra
Abstract
One of the nonparametric regression methods that can be used for classification is
Multivariate Adaptive Regression Splines (MARS) which is enhanced using bootstrap
aggregating (bagging) with 50 replications. This method is applied to conventional crime
data, namely cases of theft which can be seen based on crime rates in Riau Province in
2016-2020. The dependent variable used is the theft crime rate, while the independent
variables are population density (𝑋!), poverty rate (𝑋"), RLS (𝑋#), and PDRB (𝑋$). This
study aims to form the best model and see the results of the classification based on the
factors that influence the crime rate indicators in Riau Province. Bagging MARS method
with training data of 68% produces a minimum GCV value is 0.08961, while the MARS
method is 0.13993 in obtaining the best model. The MARS method yields 60% for
accuracy, 80% for sensitivity and specificity 40%. The best accuracy value is 85% with
sensitivity is 100% and specificity is 70% using bagging MARS with testing data by 32%.
The most influential variable using the MARS method and bagging MARS on the crime
rate indicator of theft cases in Riau Province in 2016-2020 are the poverty rate (X")
variable with an importance level of 100%.
Description
Keywords
Crime rate indicator of theft cases, classification, multivariate adaptive regression splines, bootstrap aggregating
Citation
Perpustakaan