Page 37 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

discover vehicle-pedestrian crash patterns using eight years of Louisiana crash data (2004-2011) and found the roadway lighting
at night, gender and age of pedestrians as effective parameters on crash severity [5].
In follow of previous literature, this paper proposes a novel method to explore the spatial and temporal associations
between characteristics of traffic accident reports and the 311 non-emergency service reports submitted by the citizens. The
paper is organized as follows: In section 2, we explain the presented methodology including pre-processing, and association rule
mining. Section 3 introduces data sources as well as the experimental results. Section 4 concludes the article with an eye towards
future works.

2. Methodology
To determine spatial and temporal relationships between crash characteristics and the 311 reports, association rule mining
was used. Association rule is a prominent data mining technique to extract interesting correlations among different sets of
attributes in datasets, which is used to identify the hidden patterns and rules between crashes and some urban parameters or
citizen’s report [8].
The proposed mechanism to determine spatial and temporal relationships in accident and environmental reports is shown
in Fig 1. The presented approach follows the process of three main steps in the mining process: In the first step, some
geoprocessing activities have been done to prepare the datasets for the analysis. Afterward, the Apriori algorithm is used to
explore patterns and association rules in the dataset. In the last step, the determined rules will be used to predict the probability
of accruing a traffic accident after receiving some related citizens’ reports and in some specific circumstances (i.e. land-use type,
day of week, time in a day, road and vehicle type).

Fig. 1. The workflow of determining spatial and temporal relationships in accident and the 311 environmental reports

2.1. Pre-processing
The motivation behind the pre-processing step is to prepare data for the association rule mining. The datasets are crash
reports, the 311 reports, land use data and road networks. In the first step of pre-processing, missing values, noisy data and
outliers are eliminated. To determine the related crash reports and citizen’s reports, some spatio-temporal analysis should be
done as follow:
Temporal correlation: Each report in the 311 non-emergency database has an open and a close time that represents the
submit time and the time which a reported problem has been resolved, respectively. Considering the open and close times of a
report, the temporal relationship of a report and accident type would be investigated. Therefore, the relevant reports to the
accidents are determined based on Eq. (1).

2