Using classification and regression tree and dimension reduction in analyzing motor vehicle traffic accidents

Yu Huei Liu, Kuang Yang Kou, Hsin-Hung Wu, Ya Chi Nian

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This study applies classification and regression tree (CRT) to identify the hidden knowledge in fatal accidents of motor vehicles from Fatal Traffic Accident of National Police Agency, Taiwan. In the beginning, twenty four variables are chosen from Fatal Traffic Accident data set. Later, dimension reduction is used to reduce the number of variables from twenty four to nine variables by principal component analysis. With two different CRT models with twenty four and nine variables to forecast injury severity, a comparison is made in terms of rules generated, model accuracy, type I and type II errors, and evaluation chart generated by IBM SPSS Modeler 14.2. The results show that the CRT model with dimension reduction outperforms the CRT model without dimension reduction almost in every category except for type II error since this model tends to slightly overestimate the injury severity of motor vehicle traffic accidents than the model without dimension reduction.

Original languageEnglish
Pages (from-to)639-646
Number of pages8
JournalApplied Mathematics and Information Sciences
Volume10
Issue number2
DOIs
Publication statusPublished - 2016 Jan 1

Fingerprint

Classification and Regression Trees
Highway accidents
Dimension Reduction
Accidents
Traffic
Type II error
Model
Law enforcement
Taiwan
Chart
Principal component analysis
Principal Component Analysis
Forecast
Tend
Evaluation

All Science Journal Classification (ASJC) codes

  • Analysis
  • Numerical Analysis
  • Computer Science Applications
  • Computational Theory and Mathematics
  • Applied Mathematics

Cite this

@article{06c3634ab2e34cd599fb9367803254a1,
title = "Using classification and regression tree and dimension reduction in analyzing motor vehicle traffic accidents",
abstract = "This study applies classification and regression tree (CRT) to identify the hidden knowledge in fatal accidents of motor vehicles from Fatal Traffic Accident of National Police Agency, Taiwan. In the beginning, twenty four variables are chosen from Fatal Traffic Accident data set. Later, dimension reduction is used to reduce the number of variables from twenty four to nine variables by principal component analysis. With two different CRT models with twenty four and nine variables to forecast injury severity, a comparison is made in terms of rules generated, model accuracy, type I and type II errors, and evaluation chart generated by IBM SPSS Modeler 14.2. The results show that the CRT model with dimension reduction outperforms the CRT model without dimension reduction almost in every category except for type II error since this model tends to slightly overestimate the injury severity of motor vehicle traffic accidents than the model without dimension reduction.",
author = "Liu, {Yu Huei} and Kou, {Kuang Yang} and Hsin-Hung Wu and Nian, {Ya Chi}",
year = "2016",
month = "1",
day = "1",
doi = "10.18576/amis/100223",
language = "English",
volume = "10",
pages = "639--646",
journal = "Applied Mathematics and Information Sciences",
issn = "1935-0090",
publisher = "Natural Sciences Publishing Corporation",
number = "2",

}

Using classification and regression tree and dimension reduction in analyzing motor vehicle traffic accidents. / Liu, Yu Huei; Kou, Kuang Yang; Wu, Hsin-Hung; Nian, Ya Chi.

In: Applied Mathematics and Information Sciences, Vol. 10, No. 2, 01.01.2016, p. 639-646.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Using classification and regression tree and dimension reduction in analyzing motor vehicle traffic accidents

AU - Liu, Yu Huei

AU - Kou, Kuang Yang

AU - Wu, Hsin-Hung

AU - Nian, Ya Chi

PY - 2016/1/1

Y1 - 2016/1/1

N2 - This study applies classification and regression tree (CRT) to identify the hidden knowledge in fatal accidents of motor vehicles from Fatal Traffic Accident of National Police Agency, Taiwan. In the beginning, twenty four variables are chosen from Fatal Traffic Accident data set. Later, dimension reduction is used to reduce the number of variables from twenty four to nine variables by principal component analysis. With two different CRT models with twenty four and nine variables to forecast injury severity, a comparison is made in terms of rules generated, model accuracy, type I and type II errors, and evaluation chart generated by IBM SPSS Modeler 14.2. The results show that the CRT model with dimension reduction outperforms the CRT model without dimension reduction almost in every category except for type II error since this model tends to slightly overestimate the injury severity of motor vehicle traffic accidents than the model without dimension reduction.

AB - This study applies classification and regression tree (CRT) to identify the hidden knowledge in fatal accidents of motor vehicles from Fatal Traffic Accident of National Police Agency, Taiwan. In the beginning, twenty four variables are chosen from Fatal Traffic Accident data set. Later, dimension reduction is used to reduce the number of variables from twenty four to nine variables by principal component analysis. With two different CRT models with twenty four and nine variables to forecast injury severity, a comparison is made in terms of rules generated, model accuracy, type I and type II errors, and evaluation chart generated by IBM SPSS Modeler 14.2. The results show that the CRT model with dimension reduction outperforms the CRT model without dimension reduction almost in every category except for type II error since this model tends to slightly overestimate the injury severity of motor vehicle traffic accidents than the model without dimension reduction.

UR - http://www.scopus.com/inward/record.url?scp=84960088837&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84960088837&partnerID=8YFLogxK

U2 - 10.18576/amis/100223

DO - 10.18576/amis/100223

M3 - Article

AN - SCOPUS:84960088837

VL - 10

SP - 639

EP - 646

JO - Applied Mathematics and Information Sciences

JF - Applied Mathematics and Information Sciences

SN - 1935-0090

IS - 2

ER -