=Paper= {{Paper |id=Vol-1741/jist2016pd_paper10 |storemode=property |title=Estimation of Spatial Missing Data for Expanding Urban LOD |pdfUrl=https://ceur-ws.org/Vol-1741/jist2016pd_paper10.pdf |volume=Vol-1741 |authors=Shusaku Egami,Takahiro Kawamura,Akihiko Ohsuga |dblpUrl=https://dblp.org/rec/conf/jist/EgamiKO16a }} ==Estimation of Spatial Missing Data for Expanding Urban LOD== https://ceur-ws.org/Vol-1741/jist2016pd_paper10.pdf

Estimation of Spatial Missing Data for
Expanding Urban LOD

Shusaku Egami1 , Takahiro Kawamura1,2 , and Akihiko Ohsuga1
1
Graduate School of Informatics and Engineering,
The University of Electro-Communications, Tokyo, Japan
egami.shusaku@ohsuga.lab.uec.ac.jp
ohsuga@uec.ac.jp
2
Japan Science and Technology Agency, Tokyo, Japan
takahiro.kawamura@jst.go.jp

Abstract. The illegal parking of bicycles has been an urban problem
in Tokyo and other urban areas. We have sustainably built a Linked
Open Data (LOD) relating to the illegal parking of bicycles (IPBLOD)
to support the problem solving by raising social awareness. Also, we have
estimated and complemented the temporal missing data to enrich the IP-
BLOD, which consisted of intermittent social-sensor data. However, there
are also spatial missing data where a bicycle might be illegally parked,
and it is necessary to estimate those data in order to expand the areas.
Thus, we propose and evaluate a method for estimating spatial missing
data. Specifically, we find stagnation points using computational fluid
dynamics (CFD), and we filter the stagnation points based on popular-
ity stakes that are calculated using Linked Data. As a result, a significant
diﬀerence in between the baseline and our approach was represented us-
ing the chi-square test.

Keywords: Linked Open Data, Urban problem, Illegally parked bicy-
cles

1 Introduction
The illegal parking of bicycles have been an urban problem in Tokyo and other
urban areas since the number of bicycles owned in Japan is large. Illegally parked
bicycles (IPBs) obstruct vehicles, cause road accidents, encourage theft, and dis-
figure streets. In order to address this problem, we believe it would be useful
to publish the distribution of illegally parked bicycles (IPBs) as Linked Open
Data (LOD). For example, it would serve to visualize IPBs, suggest locations for
optimal bicycle parking spaces, assist with the removal of IPBs, and assist with
the urban design. Thus, we built the illegally parked bicycle LOD (IPBLOD)3
based on social data after designing LOD schema [1]. However, there are spatial
missing data where bicycles might be illegally parked. It is necessary to comple-
ment the spatial missing data in order to apply IPBLOD to various urban areas.
3
http://www.ohsuga.is.uec.ac.jp/bicycle/dataset.html
However, it is not satisfied merely by social sensors when collecting observation
points of IPBs.
In this paper, we propose the method for geographically expanding LOD by
estimating spatial missing data. We thought that observation points of IPBs have
spatial or geographic features common such as road width and building density.
Thus, we first simulated airflow in urban area using computational fluid dynam-
ics (CFD) and found stagnation points. Next, we collected POI data around
each of the stagnation points and calculate popularity stakes of the POIs us-
ing DBpedia Japanese. Then, we filtered stagnation points if their sum of the
popularity stakes of POIs is less than the threshold. We considered the filtered
stagnation points as estimated data.

2 IPBLOD and Related Work
We have sustainably built IPBLOD and applied them to Tokyo and other several
urban areas . We collected tweets containing location information, pictures, hash-
tags, and the number of IPBs from Twitter. Also, we collected information on
POI using Google Places API4 and Foursquare API5 . Also, we obtained bicycle
parking data and weather data from websites of municipalities. These data were
used as factor data. Finally, the collected data were converted to LOD.
Bischof et al. [2] proposed a method for the collection, complementation, and
republishing of data as Linked Data, as with our study. This method collects
data from open city data such as Urban Audit6 and United Nations Statistics
Division (UNSD)7 , and then utilizes the similarity among such large Open Data
sets on the Web. However, we could not find the corresponding data sets and
thus could not apply the same approach to our study. Therefore, we estimated
spatial missing data using CFD and DBpedia Japanese.

3 Estimation of Spatial Missing Data
We consider the flow of people to the fluid, and we find stagnation points of
areas around train stations by airflow simulation using 3D maps and CFD. Also,
we filter stagnation points using DBpedia Japanese, and we regard these filtered
points as new observation points.
3.1 Finding stagnation points using CFD
We simulated the airflow around the station using Airflow Analyst8 , which is a
simulation software run on ArcGIS. The wind direction is set as being parallel
to a road from a train station, since it is considered that people come to the
station along with the roads.
4
https://developers.google.com/places/?hl=en
5
https://developer.foursquare.com/
6
http://ec.europa.eu/eurostat/web/cities
7
http://unstats.un.org/unsd/default.htm
8
http://www.airflowanalyst.com/en/index.php
Fig. 1. Patterns of stagnation points

Baseline Proposed method
Precision 0.0559 0.129
Recall 0.161 0.393
F-measure 0.0829 0.194
Table 1. Evaluation results of both base-
line and proposed method

Fig. 2. The filtered stagnation points
around of Chofu Station

A stagnation point is a point where the velocity of the fluid is zero in the flow
field. We tried to find stagnation points using patterns in Figure 1. A black node
is a node with average wind velocity x > 0.1. The white node is the node which
is x = 0. The grey node is the node that 0 < x ≤ 0.1. In general, a stagnation
point is a white node under these conditions. However, white nodes became
buildings in our experiment. Therefore, we defined grey nodes as stagnation
points. The total accuracy of the findings of stagnation points around Chofu
Station, Fuchu Station, and Shinjuku Station became the highest when we use
pattern (j). Hence, we use pattern (j) to find stagnation points in this paper.

3.2 Filtering stagnation points using DBpedia Japanese

We found the stagnation points, but, there were many noise points. We assumed
that bicycles tend to be parked illegally at stagnation points having nearby POIs,
whose popularity stakes are high. Therefore, we calculated the popularity stakes
of the POIs around of the stagnation points and then filtered the stagnation
points using the popularity stakes.
We first obtained the POIs data within a 20-meter radius from the stagna-
tion points using Google Places API. Then, we calculated the number of links
from person resources to POIs on DBpedia Japanese. All types of POIs were
manually mapped to resources of DBpedia Japanese. We considered the num-
ber of inbound links from person resources as the popularity stakes, and we
obtained the number of links from instances of foaf:Person to types of POIs.
For example, the popularity stakes of the bar resource of DBpedia Japanese
(http://ja.dbpedia.org/resource/居酒屋) became 47. Then, we calculated the
sum of the popularity stakes of POIs, and we filtered stagnation points if the
sum of the popularity stakes is less the threshold. We set the threshold to 200.
Figure 2 shows the results of the filtering. Red markers are observation points
of illegally parked bicycles. Blue circles are estimated points.

3.3 Evaluation and Discussion
We carried out the experiments on Chofu Station, Fuchu Station, and Shinjuku
Station which have multiple observation points of IPBs. The total number of
observation points was 56. The baseline estimates the spatial missing data at
regular intervals, as many as the number of stagnation points. Table 1 shows
the accuracy of both the baseline and the proposed method. As the result, the
precision, the recall, and the F-measure of the proposed method became higher
than the result of the baseline. Therefore, there is the utility of the proposed
method. Also, we validated the utility of the proposed method using the chi-
square test. The null hypothesis is that there is no diﬀerence between the result
of the baseline and the result of the proposed method, and we used a standard
level of significance p < 0.05. As the result, the p-value of precision was 7.393e-
05, and the p-value of recall was 2.244e-06. Hence, we found that there is a
significant diﬀerence between the result of the baseline and the result of the
proposed method.
The accuracy of the estimated data in this study was low for the following
reasons. The number of observation points was less. There is a possibility that
new observation points are found around the estimated points.

4 Conclusion and Future Work
In this paper, we described geographically expansion of IPBLOD by estimating
the spatial missing data. The mainly technical contribution is the proposal of
a hybrid method using CFD and DBpedia Japanese for estimating the spatial
missing data in LOD. In the future, we will estimate spatial missing data in
more urban areas, and we will check true-false results to go to estimated points.
Furthermore, we will visualize estimated observation points and will design in-
centive for social sensors (workers of crowdsourcing), in order to collect more
data related to IPBs.
Acknowledgments. This work was supported by JSPS KAKENHI Grant
Numbers 16K12411, 16K00419, 16K12533.
References
1. Egami, S., Kawamura, T., Ohsuga, A.: Building Urban LOD for Solving IPBs
in Tokyo. In:Proc. The 15th International Semantic Web Conference, pp.291-307
(2016)
2. Bischof, S., Martin, C., Polleres, A., Schneider, P.: Collecting, Integrating, Enriching
and Republishing Open City Data as Linked Data. In: Proc. The 14th International
Semantic Web Conference, pp.57-75 (2015)