Determining the Degree of Driver's Responsibility for Car Accidents by Using Data Mining Methods: The Case of Addis Ababa Traffic Control And Investigation Department
No Thumbnail Available
Date
2009-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Road traffic accidents (RT As) are now becoming one of the leading public health
problems in Ethiopia. Due to accidents people are dying and getting di sabled further
resulting in property losses. It is more severe in the capital city of Ethiopia, Addis Ababa
where most cars in the country are bustling. Most accidents in the capital are due to
driver problems. And most of the fatalities are attributed to the accident type pedestrian
hit by car.In this research an attempt has been made to apply the detIsion tree and multilayer
perceptron (MLP) neural network data mining techniques to analyse the accident data.
The research focuses on predicting the degree of dri ver' s· responsibility for car acc idents
and identifying the important factors influencing the different levels of responsibility by
using the RTA dataset of Addis Ababa Traffic control and investigation department
(AARTCIO).In the research undertaking standard data m1l1111g methodology has been employed.
Accordingly, the domain area; in this case the traffic control system is carefully studied,
the accident dataset is carefully investigated to have a clear picture and verification of the
data. Preprocessing the data by performing data cleaning, data se lection, data
transformation and replacing missing value activities are also among the crucial activities
that have been carried out in this research.Exploratory data analysis (EDA), descriptive modeling, predictive modeling
(c lassification and regression), discovering patterns and rules and retrieval by content.
In accomplishing the data mining task a number of standard techniques and tools can be
employed. The major data mining techniques according to Berry and Linoff (2004), are
decision trees, neural networks, cluster Analysis and Statistical methods li ke, Bayesian
inference, logistic regression, log-linear models, the common techniques of cluster
analysis are :which are divi sible algoritlu agglomeration algorithm s, partitional
clustering, and incremental clustering. Association Rules, genetic algorithms and fuzzy
inference systems are also worth mentioning.
Description
Keywords
Information Science