Iso-edges for the Geovisualization of Consumptions

Catarina Mac¸

as, Pedro Cruz, Evgheni Polisciuc, Hugo Amaro and Penousal Machado

Department of Informatics Engineering, CISUC, University of Coimbra, Coimbra, Portugal

Keywords:

Isolines, Big Data, Information, Consumptions, Geovisualization, Thematic Maps.

Abstract:

Data Visualization is emerging as a tool to understand and explore data in various ways. It enables us to

interpret, synthesise, and present complex and vast amounts of information. We use Data Visualization to

represent the evolution of consumptions in 729 hypermarkets and supermarkets of the biggest Portuguese

retail company, for a time span of two years. We aim to apply an Information Visualization technique in order

to study how, through Data Visualization, we can represent, synthesize, and interpret consumptions’ data. The

geospatial data enables us to represent the consumptions in the different municipal districts and to analyze

how consumptions evolve over time. To present this data, we apply an isoline approach, introducing a new

technique called iso-edges. We also implement an interface for the exploration and analysis of the data.

1 INTRODUCTION

The number of structures to collect and save data on

retail companies is increasing, and thus the necessity

to analyze and understand this data is of the utmost

importance. Data Visualization can be an invaluable

asset in this process. The representation of geospatial

data with the goal of understanding how data evolves

through time in speciﬁc regions is an important task

for retail companies. This type of information may

assist retail companies in, for instance, stock manage-

ment, understanding seasonal variations, determining

where to place new facilities, etc.

For this project, we have access to the dataset of

729 Portuguese supermarkets and hypermarkets of the

SONAE chains, which covers the entire country. We

used all transactions made on those supermarkets and

hypermarkets from May 2012 to April 2014. Further-

more, we have access to the zip-codes of all clients

of the SONAE chains, which gives us an estimate of

the clients’ geographic position, and to the locations

of the main buildings of Portugal. This data about

the consumptions was already explored through a cal-

endar visualization which aided in the understanding

of how the consumptions evolved in the different de-

partments of the SONAE chains (Macas et al., 2015;

Mac¸

as et al., 2015). We also used the geographi-

cal data of the consumptions to explore how the cus-

tomers move from one supermarket to another (Polis-

ciuc et al., ). The main goal for this project is to apply

an Information Visualization technique to represent

how consumptions are distributed throughout Portu-

gal and how this distribution evolves through time,

thus highlighting the changes in the patterns of con-

sumption in speciﬁc times of the year, such as Christ-

mas. Additionaly, it is a ﬁrst step in the development

of a visualization tool to assist decision making re-

garding the opening, or renewal, of facilities.

To represent the geographical variations of con-

sumptions, and as an initial approach, we resort to

contour lines, i.e. isolines, which delineate areas with

similar values of consumption, thus identifying re-

gions with different consumptions. Isolines depict

consumptions that were not measured in those spe-

ciﬁc geographical points, but that were calculated in

relation to the area of collection. The use of isolines

allows us to identify, delimit and highlight geographic

areas of high and low consumption values. When

applying this technique to our data we realized that

the isoline density was low, which impaired the read-

ability of the visualization. As such we introduce a

new approach to complement the isolines, iso-edges,

increasing their density in order to promote the efﬁ-

ciency of the visualization.

The paper is structured as follows: in Section 2

and Section 3 we present a short overview of spa-

tiotemporal visualizations and thematic maps, respec-

tively; Section 4 describes the data, our approach for

the geo-representation of consumptions using isolines

and iso-edges, and the interface, which enables the

user to explore the visualization; Finally, in Section

5, we draw some conclusions.

222

Maçãs, C., Cruz, P., Polisciuc, E., Amaro, H. and Machado, P.

Iso-edges for the Geovisualization of Consumptions.

DOI: 10.5220/0005785702200227

In Proceedings of the 11th Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2016) - Volume 2: IVAPP, pages 222-229

ISBN: 978-989-758-175-5

2 SPATIOTEMPORAL

VISUALIZATIONS

Data representing time and space are found in a wide

range of domains, and can be used to ﬁnd and com-

prehend natural and social patterns, and to help make

predictions. Spatiotemporal visualization techniques

can efﬁciently organize and represent temporal ge-

ographic data sets, providing a global view of the

activities, and thus revealing overall tendencies and

movement patterns (Meirelles, 2013). The applica-

tion of these visualization techniques carries some

challenges such as their computational requirements,

caused by the usually large data sets, and the need for

a sound understanding of how information and knowl-

edge can be extracted from the data, and transposed in

the visualization (Zhong et al., 2012).

Traditionally, maps have been used as models for

spatiotemporal datasets (Meirelles, 2013). Through

the use of maps, the geospatial patterns and clusters

are generally easier to identify. Zhong et al. (Zhong

et al., 2012) groups the most important spatiotem-

poral visualization techniques in 5 different types:

(1) Timestamps and Time Labels, which are series

of events marked with date and time information.

The time labels are speciﬁc graphic variables used

in maps indicating changes. This type of visualiza-

tion is static. Minards famous graphic of Napoleons

Russian campaign is one example of this technique;

(2) Baselines, which use arrows and lines to repre-

sent the changes in the data. This technique can ap-

ply animation methods to represent the progress. Na-

talia Andrienko and Gennady Andrienko (Adrienko

and Adrienko, 2011) propose a method for spatial ag-

gregation of movement data, which can be explored

in an interactive visualization that creates legible ﬂow

maps; (3) Image Series, where timelines are used

as the basis for mapping events over time, and the

dataset is represented through series of images. Guo

et al. (Guo et al., 2006) developed a tool called VIS-

STAMP which aid in the exploration and understand-

ing of spatiotemporal patterns. Their approach in-

cludes several visualizations within which one of geo-

graphic small multiple display; (4) Space-Time Cube,

developed by H

agerstraand (H

agerstraand, 1970) is a

three-dimensional diagram with time as the third spa-

tial dimension. He explored this technique to show

life histories of people and how people interact in

space and time; and (5) Real-Time Rendering of Dy-

namic 3D Scenes, which uses real-time animated 3D

rendering techniques. Weber et al. (Weber et al.,

2009) creates an interactive simulation of 4D cities,

which is based in 3D models of an urban environment

and its evolution through time.

3 THEMATIC MAPS

Thematic maps appeared in the second half of the

seventeenth century representing data in the natural

sciences, and can be deﬁned as the representation of

attribute data on a base map. Thematic maps dis-

play qualitative or quantitative data. Their purpose

is to reveal patterns and frequencies in the geography

where they occur (Meirelles, 2013), to discover the

geographical structure of the subject, and to relate its

distribution on the map (Robinson, 1982).

The 1701 isoline map of magnetic ﬁelds by Ed-

mond Halley is considered the ﬁrst thematic map.

These maps can be divided in: dot distribution

maps (Fry, 2004); graduated symbol maps (Brewer

and Campbell, 1998); choropleth maps (Brewer and

Pickle, 2002); isometric and isopleth maps (Mellier

et al., 1988); ﬂow and network maps (Thorp, 2009);

and area and distance cartograms (Meirelles, 2013).

The ﬁrst isopleth maps depicting population den-

sities were created by N. F. Ravn in 1857. The Gen-

eral Bathymetric Chart of the Oceans records from the

International Hydrographic Organization is another

isopleth example, which uses as a visual metaphor

for coloring “the deeper or higher, the darker”. It

shows the great ocean trenches of the western Pa-

ciﬁc and Japan Sea, and has numbered contours to

improve accuracy of reading (Tufte, 1990). In 2006,

Chris Lightfoot created a Travel-time Map that rep-

resents the time needed to travel from site A to site

B (Lightfoot, 2006). The visualization is drawn on a

base map where the color of each pixel depends on

the time spent, having warm colors to indicate short

journeys and cool colors for longer journeys.

For this project, we will only focus in the isopleth

maps, more speciﬁcally, in the isolines. Isopleth maps

are composed by a set of lines or areas that show

the distribution of values that cannot be referenced to

points. These lines, depict data that were not mea-

sured in those speciﬁc points, but were calculated in

relation to the area of collections. The calculated cen-

troid of each area is considered the data point for the

line construction (Meirelles, 2013). Some examples

of this technique can be seen in maps that represent

the mean temperatures or the average precipitation

levels. In 2013, the Department of Energy & Cli-

mate Change of England created an isopleth map that

shows the actual heat demand from buildings across

England (DECC, 2013). With this map developers

and planners can perceive which areas are suitable for

the development of local heat networks. Isoline maps,

are isopleths where there is no shading. Edmond Hal-

leys 1701 map of magnetic lines is considered the ﬁrst

map to make use of lines of equal value to encode

Iso-edges for the Geovisualization of Consumptions

223

data. Nowadays, the use of isolines to represent pop-

ulation data is less popular, and the majority of iso-

metric and isopleth maps, shows natural phenomenas,

such as climate and geology.

4 GEO-REPRESENTATION OF

CONSUMPTIONS

We propose the application of the isoline technique

as a method to visualize and explore the evolution of

consumptions in the different municipal districts of

Portugal through time. We also propose a comple-

mentary method, which we designate as iso-edges, to

increase the density of information presented in the

map, improving readability.

In this project, the isolines are projected on the

Portuguese map along with the representation of

the main buildings (retrieved from the Open Street

Maps

) and the zip codes of costumers. This data pro-

vides a visual reference to the viewer, highlighting on

the map the main residential conglomerates and build-

ings

. These locations are represented in the map with

small, black circles of equal size so they do not over-

power the visualization but, at the same time, are vis-

ible and can be compared with the consumption data.

For this project, and since we wanted to analyse the

differences between the different months, we calcu-

lated the average consumption of every month from

the dataset.

To differentiate the areas with different consump-

tions we draw a set of lines to separate them: isolines.

Each one of these lines, is representative of one con-

sumption value and separates the areas with more and

less consumptions than the lines’ value. With this In-

formation Visualization technique we are able to see

the geographic map behind and to analyze the differ-

ences in consumptions.

To generate the isolines, ﬁrst we had to create a

triangulation between the centroids of all municipal

districts of Portugal. We used the Delaunay Triangu-

lation algorithm to do so (Lee and Schachter, 1980).

Ideally, to avoid cluttering and promote readability,

the isolines should have a similar distance among

them, as in the work of Bruno Jobard (Jobard and

Lefer, 1997). Considering that we want our isolines

to be evenly distributed, the isolines’ values cannot be

https://www.openstreetmap.org

It is important to notice that the areas where people

live and the areas where people shop are not necessarily the

same. In certain cases areas with low population density

correspond to business centers which are highly populated

throughout the day. As such it is necessary to convey both

types of information.

predeﬁned. As such, we deﬁned a minimum distance

and represented only the isolines that have that mini-

mum distance among them. With this ﬁrst approach,

the resulting isolines were too sparse, which trans-

lated into a small amount of information, difﬁculting

the reading of the data. To give more insight, we de-

cided to complement the isolines. As such, in addition

to the previously drawn isolines, we also draw isolines

which are not closed—iso-edges. These iso-edges,

are calculated in the same way as the isolines, but, if

the isoline has segments which are too close from the

previous drawn isolines, those edges are eliminated

until the isoline is drawable. This technique resulted

in a wide number of iso-edges, which completed the

visualization and gave more insight about the data,

specially in cluttered areas created by municipal dis-

tricts which are too close to each other.

This section is divided in three sub-sections. In

the ﬁrst, we describe the dataset and explain how we

aggregated the data. In the second sub-section, we

describe how we generated the isolines and iso-edges,

and present some results. In the last sub-section we

present the interface which enables the user to explore

the visualization.

4.1 Data

The consumptions data is retrieved from 729 Por-

tuguese supermarkets and hypermarkets of the SONAE

chains, which covers the entire country. The clients

from these chains tend to use their client cards when

shopping, to accumulate discounts and other beneﬁts.

Currently, the number of active cards is above 6 mil-

lion, which can be considered an impressive number,

specially if we take into consideration that the Por-

tuguese population is below 11 million, and that the

cards are issued by “household” and shared by the en-

tire family. For this project we used all the transac-

tions made on those supermarkets and hypermarkets

from May 2012 to April 2014. Each transaction cor-

responds to one product bought and it has properties

such as price, date, place, and time of purchase. Each

product is placed in the product hierarchy of the com-

pany, which has 6 levels and corresponds to a type of

the product. For this work, we aggregate all the pur-

chases independently of the type of product. Since

the goal of the project is to give an overview of the

consumptions differences among the months of our

dataset, we also calculated the average consumption

value for every month, depending on the municipal

district where the purchases occurred. The shapes

of Portugal and its municipal districts were both re-

trieved from the Open Street Maps.

Our data, has some inherit characteristics: ﬁrst,

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

224

the consumption volumes are not related to each

other, meaning, municipal districts close to each other

can have different consumption volumes, what causes

the consumption distribution to be irregular; and sec-

ond, the distance between municipal district centers

is not the same, some districts are close to each other

and others far apart. These two characteristics are,

in some way, related: in general, the nearer from the

coast the municipal district is, the higher the con-

sumption value is and the closest the districts are from

each other; and the nearer to the interior of Portugal

the municipal district is, the lower the consumption

value is and the farthest the districts are from each

other. Furthermore, the places where the consump-

tions are higher, occur mostly in the same areas, such

as, Lisbon, Porto and Faro. These characteristics are

deeply linked with the demographics of Portugal.

4.2 Isolines and Iso-edges

To visualize the consumption differences between the

municipal districts of Portugal we implement an iso-

line technique which delimits the areas with different

consumption values. One isoline represents one con-

sumption value. To generate them, we have to deﬁne

the values that they are going to represent. Given the

inherent characteristics of our data, and the goal of

having the isolines evenly distributed along the Por-

tuguese map, we cannot divide the space of consump-

tion values in equal ranges, as our range of consump-

tion values are not linearly distributed by the munic-

ipal districts. Furthermore, in our consumption data

there are more municipal districts with low consump-

tions then with high consumptions, so if we divide the

range of consumption values in a linear way, the iso-

lines would be very irregular, some too close to each

other, others too far apart (Figure 1).

To attain a more uniform and informative distribu-

tion of isolines over the map, we deﬁned their values

using an iterative process. First, we ﬁnd the higher

consumption value of all months, so that it is pos-

sible to compare the differences among the different

months, and use this value, iso

max

as reference for an

isoline. The remaining isolines are obtained by de-

creasing values of consumptions with a sampling step

of step, in other words, the i

isoline corresponds to a

consumption value of iso

max

−i×step. The higher the

step is, the fewer isolines are calculated, resulting in a

map with less detail, since the range of consumption

values is divided in fewer parts. In order to have as

much isolines as possible without compromising leg-

ibility, we used a step of 100, so small differences in

the consumptions are discarded. Once the isolines’

values are calculated, the drawing algorithm deter-

Figure 1: Isolines generated with a linear division of the

consumption values.

mines which isolines are drawn. We start by draw-

ing the isoline corresponding to iso

max

and proceed

by iteratively decreasing the consumption values; An

isoline is drawn if, and only if, all of its points are,

at least, at a minimum predeﬁned distance, min

dist

of the isolines that were already drawn. The isolines

which are too close to those that were already drawn

are discarded. The values for the constants step and

min

dist

are empirically determined.

Before computing the isolines, it is necessary to

create a triangulation between the municipal districts

centroids. These centroids are pre-calculated for each

municipal district shape through the centroid of the

polygon (Bourke, 1988). Afterwards, we imple-

ment the Delaunay Triangulation (Lee and Schachter,

1980) for these centroids, and we get an array of tri-

angles that link the closest municipal district centers.

Note that our triangulation does not connect the cen-

troids of the municipal districts that are in the edges

of Portugal and have more than one municipal district

centroid in the middle (Figure 2) (Ahlen, 2010). The

connection of all the centroids would create isolines

which are out of the area of Portugal and that do not

represent the data.

For every month, each centroid has a value cor-

responding to the consumption in that municipal dis-

trict. To ﬁnd the points that compose the isoline cor-

responding to a given consumption value, we search

pairs of vertices of the triangulation whose associ-

ated consumption values deﬁne an interval compris-

ing the value of the isoline. To position the isoline

point along the vertex deﬁned by these two points, we

interpolate linearly between the the vertices’ values.

Iso-edges for the Geovisualization of Consumptions

225

Figure 2: Triangulation of the municipal districts of Portu-

gal. Note that there are no connections out of the Portugal’s

Shape.

Figure 3: Six centroids [A..F] and corresponding consump-

tion values. The points of the isoline corresponding to a

consumption value of 20 are determined through interpola-

tion between the extremes of each edge.

Figure 3 illustrates this process.

Once the points of the isoline are determined, the

isoline is deﬁned by connecting the two points of each

triangle belonging to the isoline path

, generating an

iso-edge. We save each iso-edge in a hash map that

has as key the consumption value of the isoline and

as value an array of the edges. This way, we easily

separate the iso-edges of different values.

Once all the iso-edges are calculated, we group

the edges that belong to the same isoline and see if

the isolines have a predeﬁned minimum distance from

each other. To generate an isoline, and since an iso-

edge is represented by two points, we just need to join

the iso-edges that have one point in common.

We generate all isolines for every isoline value and

If the triangle belongs to the isoline path, then it will

always be intercepted in exactly two points, except when

all the vertices of the triangle have the exact value of the

isoline. In that case the isoline matches the shape of the

triangle

Figure 4: Schematic of the isoline closing technique. 1)

Projection and connection of the Municipal District cen-

troids; 2) Connection of the end of the isoline and it’s pro-

jection; 3) closing the isoline.

save them. At each iteration through the isoline’s val-

ues, we need to determine which isolines are to be

drawn. To do so, and to make this process faster, we

deﬁned a radius, so the algorithm only searches for

isolines whose centroid is close to the current isoline’s

centroid. For each point of the closer isolines, we

compare the distance among the points of the current

isoline. If the distance between two points is smaller

than the minimum distance, we do not draw the cur-

rent isoline and repeat the process for the next isoline.

It is important to note that we start our algorithm with

the isoline with higher consumptions, so no isoline

intersects. With this method we are able to create a

relatively uniform distribution for the isolines.

Next, since the triangulation does not use the

edges of the Portuguese map, it is necessary to con-

nect the isoline’s endings which are placed near the

outline of Portugal. We decided to: project the cen-

troid of the municipal district by placing it in the mid-

dle of the coast line; map the isoline value into an

imaginary line that connects the projected centroid of

the two municipal districts; generate a line that con-

nects that point to the end of the isoline; and draw

it until it reaches the outline of Portugal. This way,

there is no isoline with unclosed endings (Figure 4).

The ﬁnal step for the creation of the isolines is

the rounding of the isolines’ corners and the color-

ing. For the rounding of the isoline’s corners we ap-

plied a method described in (Polisciuc et al., 2015).

With this method we were able to smooth the corners

and to give a more organic aesthetics to the visual-

ization. With all isolines set, it is necessary to distin-

guish the different consumption values that they rep-

resent. Since our data has substantial differences in

the number of isolines with high and low consump-

tions, a linear color palette, that goes from one color

to the other to represent the range of values between

the lower and higher consumptions, would not differ-

entiate enough the low values and would create detail

in the high values, which is not desired. Therefore,

we need to create an algorithm to generate a color

palette that gives more detail in the low values and

less detail in the high values. First, we calculated all

the isolines which are drawn in each month of the two

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

226

Figure 5: Differences of the visualization with iso-edges

(left) and without (right). One can perceive that in the left

side of Portugal, specially in the center and upper part, the

iso-edges generate more density, thus creating more insight

about the consumptions.

years of data, and group them by consumption value.

Note that, having calculated the color palette for every

month enables us to use the same color palette, and

thus enables us to compare the different months and

the evolution of consumptions over time. To deﬁne

the color palette we divided our consumption values

in equal sized bins (quantile) to group the consump-

tion values in different ranges. With this method, we

have more colors to distinguish the isolines with low

consumptions, since there are more isolines in those

ranges, and less detail in higher consumptions. Our

color palette goes from the cold colors, for the low

consumptions, to the warm colors, for the high con-

sumptions.

Since some centers of the municipal districts are

too close to each other, and the isolines are restricted

to a minimum distance among them, we get few iso-

lines in some areas of Portugal. This is translated in

a lack of information about the consumptions, which

does not fulﬁll our objectives. To generate a complete

visualization of the consumptions we propose a new

method—iso-edges. This method, creates more den-

sity in our visualizations promoting readability. We

used the isolines that were not drawn in the previous

approach. Implementing a similar algorithm, every

time that an isoline cannot be drawn, we save it into an

array. When all the drawable isolines are generated,

we start again with every non-drawable isoline. Then,

to each point of the isoline we see if it is too close, or

not, to the points of the other isolines. Here, instead of

not drawing the isoline, if one or more points are too

Figure 6: Visualization of the consumptions in December

2012 (left) and January 2013 (right). Between these two

months, the average consumption is similar, having higher

differences in the center of Portugal, higher consumptions

in December than in January.

Figure 7: Visualization of the consumptions in February

2013 (1) and February 2014 (2). Small differences in the

generated isolines and iso-edges. Higher consumptions in

February 2013 in the west coast of Portugal.

Figure 8: Visualization of the consumptions in October

2012 (1) and October 2013 (2). One can perceive the open-

ing of a new store in October 2013 through the appearance

of new isolines.

close, we cut the isoline in edges, so that the drawable

edges have, at least, a minimum distance from the

other isolines already drawn. Figure 5 highlights the

differences between including or not the iso-edges.

In Figure 6 we can see the results of our approach.

With the use of the color palette we can better distin-

guish the areas with different consumptions. In Fig-

Iso-edges for the Geovisualization of Consumptions

227

Figure 9: Isolines interface.

ure 7, we can perceive that the consumptions were

higher in 2013 than in 2014 in the region of Lisbon.

It is possible to see that our isolines do not have sig-

niﬁcant changes as time passes. They are almost in

the same locations only with different colors. This is

due to the fact that the triangulation is the same and

the places where the consumptions occur also stay the

same. The only exception to this, is the opening of a

new store in October of 2013 in the district of Sines.

In the visualization (Figure 8), it is possible to ﬁnd a

new set of isolines in the south of Portugal (2).

4.3 Interface

To facilitate the interaction and exploration of the vi-

sualization, we developed a intuitive interface which

allows the user to explore the map and to change

the displayed time period (Figure 9). To change the

month, the user has a slider on the right side of the

interface. This slider allows the user to choose be-

tween the different months simply by clicking on the

intended month. The user can also interact with the

map by zooming, using the buttons in the right upper

corner or by using the scroll of the mouse. The zoom

has different levels in which the isolines and iso-edges

are updated. As the user zooms in, more isolines and

iso-edges appear creating a more detailed map of the

consumptions in the zoomed area (Figure 10). The

user can also drag the map and explore different areas

of Portugal.

In order to have a more extensive understanding

of the consumptions, it is also possible to see the lo-

cation of the stores where the consumptions occurred.

To do that, the user has to click in the upper right cor-

ner, in the “lojas” (stores) toggle. Finally, to help the

user to read the values, a caption with the color palette

of the isolines is displayed in the right bottom corner.

5 CONCLUSION

Through Data Visualization we are able to explore,

comprehend and synthesize big and complex datasets.

With access to the geo-localization of the hypermar-

kets we represented the evolution of consumptions in

Portugal in the 729 hypermarkets and supermarkets of

the biggest Portuguese retail company, SONAE. We

applied an isoline technique to differentiate the ar-

eas with different consumptions. An isoline is a line

which represents a value that is calculated through the

consumption value in the municipal districts. Our ﬁrst

approach was considered unsatisfactory since, due to

the dataset characteristics and the demographics of

Portugal, the resulting visualization conveyed little in-

formation regarding the areas of Portugal with high-

est population densities and consumptions. As such,

we introduced a new technique, iso-edges, which can

be deﬁned as unclosed isolines. This technique over-

comes the problem by providing additional visual in-

formation to these regions, which are, arguably, the

most important ones from a facility allocation stand-

point. With this technique, we are able to analyze the

differences in consumption among the municipal dis-

tricts, to analyze the areas which are not covered by

the SONAE’s hypermarkets, and to highlight areas of

low consumptions and high populations, which can be

promising locations for opening new stores. Further-

more, we implemented an interface to enable the user

to easily explore the visualization. Through this inter-

face the user can access different zoom levels which

give to the visualization of different levels of com-

plexity.

Future work will focus on providing, through vi-

sualization, a tool to further support decisions con-

cerning the opening of new facilities. To this end we

will take into account the number of inhabitants per

region, their estimated income, the percentage of ac-

tive and inactive costumers, the proximity to existing

surfaces and their capacity.

ACKNOWLEDGEMENTS

This project was developed within a partnership with

SONAE: Sonae Viz – Big Data Visualization for retail.

REFERENCES

Adrienko, N. and Adrienko, G. (2011). Spatial generaliza-

tion and aggregation of massive movement data. Visu-

alization and Computer Graphics, IEEE Transactions

on, 17(2):205–219.

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

228

Figure 10: Different zoom levels.

Ahlen, T. (2010). Poly2tri, a 2d constrained delaunay trian-

gulation library. https://code.google.com/p/poly2tri/.

Bourke, P. (1988). Calculating the area and centroid

of a polygon. URL. http://local.wasp.uwa.edu.au/ -

pbourke/geometry/polyarea.

Brewer, C. and Campbell, A. J. (1998). Beyond gradu-

ated circles: varied point symbols for representing

quantitative data on maps. Cartographic Perspectives,

(29):6–25.

Brewer, C. A. and Pickle, L. (2002). Evaluation of meth-

ods for classifying epidemiological data on choropleth

maps in series. Annals of the Association of American

Geographers, 92(4):662–681.

DECC (2013). National heat map shows heat de-

mand from buildings. http://infosthetics.com/

archives/ 2012/03/national heat map shows heat

demand from buildings.html.

Fry, B. (2004). Zip code map. http://benfry.com/zipdecode/.

Guo, D., Chen, J., MacEachren, A. M., and Liao, K. (2006).

A visualization system for space-time and multivari-

ate patterns (vis-stamp). Visualization and Computer

Graphics, IEEE Transactions on, 12(6):1461–1474.

agerstraand, T. (1970). What about people in regional sci-

ence? Papers in regional science, 24(1):7–24.

Jobard, B. and Lefer, W. (1997). Creating evenly-spaced

streamlines of arbitrary density. Springer.

Lee, D.-T. and Schachter, B. J. (1980). Two algorithms

for constructing a delaunay triangulation. Interna-

tional Journal of Computer & Information Sciences,

9(3):219–242.

Lightfoot, C. (2006). Travel-time maps. https://www.mysoc

iety.org/2006/03/04/travel-time-maps-and-their-uses/

Macas, C., Cruz, P., Amaro, H., Polisciuc, E., Carvalho, T.,

Santos, F., and Machado, P. (2015). Time-series ap-

plication on big data – visualization of consumption in

supermarkets. In Braz, J., Kerren, A., and Linsen, L.,

editors, IVAPP 2015 – Proceedings of the 6th Interna-

tional Conference on Information Visualization The-

ory and Applications, Berlin, Germany, 11-14 March,

2015., pages 239–246. SciTePress.

Mac¸

as, C., Cruz, P., Martins, P., and Machado, P. (2015).

Swarm systems in the visualization of consumption

patterns. In Yang, Q. and Wooldridge, M., editors,

Proceedings of the Twenty-Fourth International Joint

Conference on Artiﬁcial Intelligence, IJCAI 2015,

Buenos Aires, Argentina, July 25-31, 2015, pages

2466–2472. AAAI Press.

Meirelles, I. (2013). Design for Information: An Introduc-

tion to the Histories, Theories, and Best Practices Be-

hind Effective Information Visualizations. Rockport

publishers.

Mellier, Y., Mathez, G., Mazure, A., Chauvineau, B., and

Proust, D. (1988). Subclustering and evolution of the

coma cluster. Astronomy and Astrophysics, 199:67–

72.

Polisciuc, E., Alves, A., and Machado, P. (2015). Under-

standing urban land use through the visualization of

points of interest. In Proceedings of the Fourth Work-

shop on Vision and Language, pages 51–59, Lisbon,

Portugal. Association for Computational Linguistics.

Polisciuc, E., Cruz, P., Amaro, H., Mac¸

as, C., Carvalho, T.,

Santos, F., and Machado, P.

Robinson, A. H. (1982). Early thematic mapping in the

history of cartography.

Thorp, J. (2009). Just landed: Processing, twitter, metacarta

and hidden data. http://blog.blprnt.com/blog/blprnt/j

ust-landed-processing-twitter-metacarta-hidden-data

Tufte, E. (1990). Envisioning information. Cheshire

(Conn.), page 91.

Weber, B., M

uller, P., Wonka, P., and Gross, M. (2009). In-

teractive geometric simulation of 4d cities. In Com-

puter Graphics Forum, volume 28, pages 481–492.

Wiley Online Library.

Zhong, C., Wang, T., Zeng, W., and Arisona, S. M. (2012).

Spatiotemporal visualisation: A survey and outlook.

In Digital Urban Modeling and Simulation, pages

299–317. Springer.

Iso-edges for the Geovisualization of Consumptions

229