S-PLOR-T: An Interactive Urban spatio-temporal Data Exploration Web

Approach
Miguel A. Feijóo* José T. Hernández†
Universidad de los Andes Universidad de los Andes

A BSTRACT multidimensional spatio-temporal cube, where the Wh* questions

The analysis of spatio-temporal urban data has evolved along the last are put. For example, when knowing the What and Where, an
30 years. Data heterogeneity and segmented analysis is a frequent analyst would probably desire the When; when knowing the When
issue in this context. We present a new proposal to address these and What, an analyst would probably want to know the Where,
challenges by using the visual analytics approach in a web environ- among others [2]. Moreover, this study provides a palpable case
ment using the “What-Where-When“ pyramidal paradigm to drive study that seeks to familiarize users with the tool in graphical and
the analysis. Thanks to the flexibility of new web tools, we arrive analytical terms, by the fulfilling of challenges previously posed
to offer with S-PLOR-T a flexible and multi-screen web interface. with different levels of complexity, thought in a specific domain
Furthermore, we provide an architecture to allow to put coordinated knowledge for analysis (urban transportation).
visualizations, letting an analyst to have a clear notion of its data
2 R ELATED W ORK
through interactions within the three main “Wh*“ dimensions. With
S-PLOR-T, the users will result in a customizable analysis in or- Towards the analysis and exploration of big amounts of data that is
der to support the finding of new insights and support an effortless usually yearned, a visualization widget for summarizing, exploring
decision making with them. Additionally, our approach allows the and navigating multivariate datasets was implemented to achieve
user, not only to visual query any view, but to explore and to analyze this current challenge. Guerra introduces and evaluated the tool
multi-scale attributes within the space and the time. In this paper, we called ”NAVIO”, which displays full summaries, allows sorting
present a use case which was built and tested with domain experts and filtering on ranges and values, and keeps a visual trail of the
in land use and mobility. queries, allowing users to navigate and explore the data effortlessly.
As Guerra mentions in his paper [11], Summarization, referring to
Index Terms: Human-centered computing; Visualization tech- getting a general idea of the dataset as a whole, Navigation, and
niques; Visual analytics; spatio-temporal data; Multivariate analysis Exploration, involving the availability to run specific desired queries,
are the three main tasks that Navio addresses.
1 I NTRODUCTION Among a large group of commercial tools that support visual
Nowadays, data analytics is turned to be the future of management analysis, such as Infozoom or Tableau, the latter is the best known
of various activities in the industry and academics. However, bar- and common to achieve these visual tasks easily. Due to its nature, is
riers on becoming more data driven for organizations or analysts the user who decides, among the dimensions or values, to visualize
are generally based on managerial and cultural rather than related dragging and dropping each element on a canvas. Being a common
to data and technology. That is, the leading obstacle in terms of commercial tool, the tool is intuitive when you have knowledge of
visual analytics’ adoption is the understanding on the alternatives of what you are looking for, emphasizing its advantages over navigation
analysis to carry their needs (e.g lack of skills or general contextual tasks, supporting big amounts of data and processing them with a
knowledge of specific data or analytics‘ alternatives), as also costs high efficiency.
and invested time, knowing the advantages or benefits of carrying Current works with variety of temporary query tools such as
out of this, though. [12] TEMPEST [1] allows the user to select arbitrary combinations of
Thus, through the development of this project, is intended to let months, days and times of day and see what happened at this time.
the users to result in a customizable analysis to support the finding Furthermore, another type of interactive time filter in the STRAD [9]
of new insights in a web approach thought to be unlicensed and project where the user selects the period of time that he wishes to
currently without pricing for exploratory analysis terms to data with see by means of a slider where the consultation period is specified,
such spatial and temporal characteristics. Additionally, allowing the placing the start and end date and time, allowing to visualize on the
user, not only to visual query any view, but to explore and to analyze map the trajectories of ships that moved during that time. Moreover,
multi-scale attributes within the space and the time. With this web TrapezoidBox [4] is a reference tool for the spatial query, imple-
approach the alternatives of accessibility to data analytics will be mented on Google Maps, in which the spatial proximity queries are
fulfilled in such way that the users are permitted to find insights made by means of a trapezoid, representing the ranges satisfactory
with such coordinated analysis within the multivariate, spatial and distance from the place of interest. Likewise, ”ANIMAP” [8], which
temporal attributes in a much more accessible manner. support the comparison through scenarios and prediction in terms of
Furthermore, by providing coordination features between views both spatial and temporal analysis with the immersing of the simu-
within the framework of the What, When and Where pyramidal lation model MARS (Metropolitan Activity Relocation Simulator),
paradigm, the project provides different analysis‘ alternatives that for Transport Planning.
imply various attributes, such as the temporal or the spatial ones. By its part, Andrienko proposes a topology based on the classi-
The latter, as representing the multivariate analysis through the fication and evaluation of how such spatio-temporal data can help
to resolve questions through exploration and the characteristics of
* e-mail: [email protected] the data they are applicable to [2]. The proposal directly relates
† e-mail: [email protected] tasks to components of data, denoting the (1) which: Identification
Attributes, (2) the when: Temporal attributes, (3) the where: Spatial
attributes and (4) the what: Descriptive attributes.
On the other hand, Chen et al. (2018) proposes a Visual Analyzer
for Urban Data called VAUD which supports the visualization, query-
ing, and exploration through multi-sources permitting the extraction to what is called cross-filtering centered in the data by itself. That
of information that would be hidden to a single data subset [6]. Co- is, through the interaction over any module, adjusting over the other
incides with Andrienko when defining the resolution of a question modules displayed will be made through linked filtering, even if the
on a query as the combination of four main component constraints, representation of the data is different between its views.
also claiming the importance of visual querying, to engage more Additionally, the application consists of graphic components that
non-expert users. depend on the interaction provided by the user, which is the involve-
Furthermore, Doraiswamy et al. (2018) states that visual analytics ment of logical components. That is, the elements presented in the
systems such as their tool proposed as URBANE, aim to empower presentation layer will depend on the communication carried out
domain experts to explore multiple data sets, at different time and between the different modules that are displayed according to the
space resolutions [7]. In this proposal navigation and operations discretion of each user. In this way, the presentation layer consists
on map view such as panning, zooming, and rotating the view are of “n“ number of modules/containers that will represent information
accomplished through mouse interactions for analyzing multiple provided to the user (Summary/Statistics/SpatialTemporal Analysis,
sources. From all the above, generally, some of the solutions in the among others). Moreover, the architecture supports the aggregation
same domain of S-PLOR-T have a limitation in the synchronized of new modules, which gives the characteristics of S-PLOR-T to
analysis between the pyramidal paradigm (WH *), which is rarely be as a toolbox where can be added or removed different analytical
met in current tools . On the other hand, none of the solutions tools, in development terms in an effortless manner. On the other
resulting from this revision, within this spatio-temporal framework, hand, the main axis of communication lies in a parent module that
do not comply with the scope of the web. However, it is important is the Data Manager, from its data. Through a event hub, all mod-
to mention that each of the tools in this comparison framework ules listen and receive constantly to the data provided by the data
meets the objectives for the contextual objective for which they were manager of the presentation layer.
designed and implemented, with specific functionalities in their Now, speaking in logical terms, from the diagram’s layer below
domain. the Presentation‘s, can be seen three well differentiated parts, where
the components use the services to obtain the data to be presented in
3 S-PLOR-T: T HE TOOL / D ESCRIPTION the templates.
A web application approach is proposed and shown within the ex- The latter is represented in a web environment based on views /
ploration of urban data, with temporal and spatial features, that lets containers coordinated through data from a mother module (Data
a user have a general notion of what is represented in the set an- Manager), for integrated analysis and set framed within the spatio-
alyzed. Everything turns around a paradigm that expresses three temporal pyramidal paradigm (WH*), as presented in Figure 2, rep-
main dimensions asked when having data sources with these char- resenting the analysis set with the data charged to the web framework
acteristics. Therefore, following the pyramidal paradigm, the re- posed.
lation between two WH* question will help to answer the third
remaining. Thus, the visual solution proposed carries different
containers/modules that will be displaying linked views within the
framework of the paradigm presented previously. Respecting the
Shneiderman’s mantra for Visual Analytics [13], is pretended that
the tool provides the (1) overview first, (2) zoom and filter, and (3)
details on demand features, as the user requires for a specific and
perhaps unique analysis.

Figure 2: S-PLOR-T: Tool‘s overall specification

4 S TUDY C ASE : “B EHAVIOR OF DAILY URBAN TRIPS BY

HOME - S TUDY CASE IN B OGOT À CITY, C OLOMBIA “
Within the framework of the master‘s thesis in Civil Engineering
developed by the researcher Javier R. Peña, from Universidad de
los Andes, it is proposed as a brief case study, for this particular
project to use a sample of a set of provided data for this investigation.
The group: “Group of studies on urban and regional sustainability -
SUR“, which is a working research group of the Universidad de los
Andes, of the academic unit: Department of Civil and Environmental
Engineering, provided the data from the Household Travel Survey
of Bogotà for the year 2015, which states: “... the most complete
official source of information on mobility and travel patterns ...“.
Figure 1: Architecture followed by the tool: S-PLOR-T (Component This survey is the compilation of the one-day travel diary of peo-
Diagram). ple over 5 years of age living in participating households located
in the area that includes the city of Bogotà and its 17 neighboring
As the architecture described in Figure 1, it permits to have some municipalities. [14] This provides trip information of two points
flexibility over every physical container displayed. That is, the reor- (origin-destination) per member per household. In addition, sociode-
ganization on the same multiple views over a single web browser, mographic information is also included, such as age, socioeconomic
maximizing, minimizing or even closing them at the user‘s conve- stratum, means of transport, gender, etc.
nience, as well as setting a single or set of containers in multiple The analysis of cities such as Bogotá, as in terms of urban mobil-
web browsers. Therefore, a user will have the opportunity to decide ity, allows finding, for example, patterns that allow determining the
how and what to have displayed for a specific analysis. Moreover, level of development of the city in terms of transport, considering
the link between every container is explicitly or implicitly applied the nature of distribution of households in the urban space.
 Thus, getting to know those patterns of mobility of great amount in order to close with the respective conclusions seeking to answer
of people in the city, may tell Bogotá’s planners to make quick deci- the question posed for the present study.
sions just by obtaining descriptive analysis, as spatial and temporal It is important to emphasize that each participant, following the
analysis, in an effortless but objective manner. elements and times provided spent in average one (1) hour in the
user test.
5 E XPERIMENTAL P HASE
5.2 Experimental Results
To evaluate the proposed tool in graphic terms, of interaction and
according to the perception of it by real users, an experimental pro- In the first hand, in terms of time spent by all participants there is
tocol was defined and designed for that purpose. In a period of one an interesting relationship that indicates the versatility of each of
(1) week, user tests were carried out among 47 people who volun- them with the tool as the challenges posed (five) were passing. The
teered. The totality of the participating users was divided into two foregoing, as results of the familiarization of the tool and its various
study groups of interest: (1) Undergraduate course ”Transportation functionalities to support the desired analysis. This can be evidenced
Systems” (7th Semester) - Civil Engineering, Universidad de los in detail in Figure 3 where, as the challenges progressed, the time
Andes and (2) Students of Master‘s degree and higher educational invested for the resolution of them was decreasing in significant
rank of the Research Group: ”Group of Studies on Urban and Re- terms.
gional Sustainability - SUR ”- Civil Engineering, Universidad de los
Andes. The first considered as ”Apprentices” and the second group
as ”Experts”.

5.1 Experimental Methodology

In the first hand, every participant was introduced to the proposed
tool, showing them the generalities of every module/container sep-
arately. Through this exercise the participants of the study were
immerse on what the main and general functionalities every compo-
nent of every container has. This last, in order to contextualize them
and letting them know how they could get results from analysis from
different and perhaps linked perspectives.
Then, each participant presented an Usability Evaluation in an
individual manner. This exercise pretended to evaluate, per user,
its understanding on the use of the tool to reach the solution of a
quantity determined of challenges, that let to interact with various Figure 3: Invested times per challenge per participant
components of the tool as a whole. That is, every challenge per-
mits the interaction with a particular set the attributes of the views Notwithstanding the latter, it should be emphasized that the par-
proposed to get to the answer of it. Additionally, was asked about ticipating groups differ in the time invested per challenge. The
YES/NO questions that are related to the completion by challenge foregoing, due to possibly to the difference of contextual knowledge
and how it was perceived to be in terms of difficulty. Moreover, com- in terms not only of alternative tools to supply specific analyzes, but
ments per challenge are also asked, so that the perception received also in terms of the data used for that purpose. This is a consequence,
has more context and description. Time was set and calculated per preliminarily concluding, of the difference in the educational level
challenge per user. All the above, with the purpose of determining that differentiates both groups, result of perhaps the interest in the
the answers got and how were them perceived in general terms. proposed analysis or generic knowledge in analytical terms in the
Moreover, for each challenge posed, the invested time using the tool statistical (multivariate), spatial or temporal framework.
was calculated per participant. On the other hand, in terms of the analysis of study variables
The challenges posed are defined considering the participating of interest, they are represented in terms of (1) Completeness, (2)
groups‘ domain knowledge, for which each was defined to get them Difficulty and (3) Correctness. In this way, in a percentage way
interact with various components of the tool within the pyramidal we want to determine by the behavior of the users with the tool if
paradigm, as follows: “Consider the case in which it is desired to they considered the proposed analysis to be complete, regardless of
know the general proportion of the predominant means of trans- whether their response to each is correct or not, if they considered the
portation of the people whose reason for traveling is working at resolution by challenge difficult, and an analysis of the percentage
night (considered as from 6:00 PM to 6:00 AM). Additionally, is of how many got their answers correct in the expected responses for
required to know in average the stratum, its neighborhood and each challenge.
locality, and cardinal point in the urban space of the city (North-
South-East-West), where these people are mostly concentrated using
the two most frequented means of transportation, from the previous
analysis.”
After the above, once the study is completed, an evaluation of the
quality of the tool is made by each of the participants of the test, to
indicate the positive and negative aspects that they have experienced,
with a view to improving the first version of the tool. It was posed a
6-position scale evaluating the participants‘ perception about ease of
use, flexibility and the compliance with tool‘s objectives from which
it was designed and developed, in both analytical and graphical
terms, where the results from 0 to 2 are considered as negative results
and from 3 to 5 as positive. The latter, besides the sociodemographic
information also asked to each participant. Figure 4: S-PLOR-T: Study Variables‘ Results by Participating Group
Thus, having the results obtained with the participation of the
users of this study, we proceed to carry out the due analysis of these, From the results presented in Figure 4, is evident how the standard
deviations for each study variable is bigger for the apprentices than
the expert users. Following the same previous idea on the invested
times per challenge, this might be due to the expertise that the users
have got in analytical terms. That is, there is a higher variation
in all the results, turned to be mid-term of low-term results for
the inexperienced users, considering the level of maturity framed
by the contextual knowledge not only of analysis‘ alternatives but
Figure 6: S-PLOR-T: Interest and compliance of the objective on the
of the data with which the users were working during the user tool
tests. However, it is worth to say that the expert users coincide in
completing all the challenges posed, in an similar average time as
they were acquiring versatility over the tool during the passing of the
able to show how this tool does meet the objectives for which it was
challenges, considering the resolution of them without difficulty or
designed and implemented. The above is represented in the right
a high effort, and getting the answers expected to be as correct. That
hand of Figure 6 where the percentage indicated by the totality of the
is, as the expert users have a complete contextual knowledge and a
participants is higher than 85% according to the general perception
high maturity level in analytical terms, all the results in comparison
received. The above in favor of answering the main research question
with the apprentices were significantly positive. The latter, letting
that frames this project, which requires affirming or denying whether
us conclude in a preliminary way that this kind of tools, such as S-
a web approach for multidimensional visual analysis through syn-
PLOR-T, requires a deep understanding towards the analysis desired
chronized views facilitates the analysis of spatio-temporal data in
over a specific known data and its attributes, in order to get the most
terms related to the paradigm of what, when and where, eliminating
out of the tool as a facilitator of exploratory data analysis within the
common barriers, such as accessibility, availability, costs, licensing
paradigm of what, when and where.
restrictions, among others, for users/analysts.

6 C ONCLUSIONS
From all the experimental methodology carried out, the posed tool
S-PLOR-T does meet the main objective of being a facilitator for
exploratory data analysis with spatio-temporal characteristics in a
flexible and user-friendly manner, with a contextualized approach
towards Visual Analytics. The latter due to the coordination between
views through the three (3) dimensions (WH*) in a web environment
letting its service 7/24, its flexibility on the viewing surface (Visual
“Lego”), its flexibility in development terms as being an integrator
of new tools for analytics, being a tool as big as the analysts require,
among others. This study showed that a web scope for the flexi-
ble and pleasant development of visual analytics within the linked
paradigm of what, when and where is a good alternative for users
who require easy accessibility to agile exploratory analysis.
On the other hand, all the results differ among the participating
Figure 5: S-PLOR-T: Likert 6-scale analysis for Perception groups, which allows the inferring that the level of maturity in
data analytics (level of expertise) is fundamental to overcome the
Furthermore, in terms of the general perception in the framework learning barrier of the tool. Additionally, is evident how there is
of ease of use, flexibility and compliance with the objective by which a high relationship between the results obtained in the challenges
the tool was designed and developed, as presented in Figure 5, fol- (time spent, correctness, completeness, among others per challenge)
lowing the same natural behavior of the usability evaluation, both per participant with their perception of the tool S-PLOR-T. From
participating groups differ, perhaps representing a high relation be- all the comments made in general, is shown how they indicated
tween all the study variables defined and the perception received by that S-PLOR-T facilitates the time that is spent in the analysis in
the tool, particularly by each of the participating groups, generically comparison with other known tools used by them, for its accessibility
speaking. This can be evidenced in how the experts considered this a as being Web and its complete synchronization between several
flexible and easy-to-use tool in terms of both graphical and analytical windows (containers). Notwithstanding the latter, is needed a prior
terms, within the spatio-temporal framework. However, in a similar contextual knowledge of the data and the tool, to get the most out of
way to inexperienced users or apprentices, they agreed to identify the analysis tool provided.
improvements in the tool, for future versions of this project, in terms Considering the analytical process of this study, all the results
particularly of the ease of locating the content and functionalities were analyzed in a descriptive manner, due to the scope of this
of the tool. The above, considering the acquired versatility scheme project willing to analyze the participants’ behavior and perception
as the user becomes familiar with the tool. That is, the process of the tool, principally. Notwithstanding the latter, there are much
of breaking the learning barrier in terms of the use of the tool is more sophisticate alternatives to get a deeper analysis (e.g. T-Tests,
based on the previously acquired contextual knowledge in analytical Eye Tracking as standardized protocol for User Experience, among
(e.g. statistical) processes, such as the interaction with the different others) [3, 10], that were not considered preliminary for analysis.
sections of the tool and the use of the various functionalities for Thus, from all the results obtained in general terms in this pre-
analysis each user win individually. liminary study, S-PLOR-T has great opportunities in the future for
Notwithstanding the latter, is interesting as in general terms for being a great web approach for data analytics, with such features that
the 47 participating users a percentage higher than 90% coincided permit the analysis centered on the paradigm of the WH* question,
in determining the S-PLOR-T tool as ”interesting”, as indicated by as incrementally as required.
the left hand of Figure 6 which gives an indication of interest and
liking for the tool as an integrating project of the paradigm of what, ACKNOWLEDGMENTS
when and where in a single solution. On the other hand, above the The authors thank to Javier Peña (M.Sc.), professor Luis Ángel
interest in terms of perception given by users, it is interesting to be Guzmán (PhD.), and research group: SUR, for all their big support.
R EFERENCES
[1] N. Andrienko and G. Andrienko. Interactive visual tools to explore
spatio-temporal variation. In Proceedings of the working conference
on Advanced visual interfaces - AVI '04. ACM Press, 2004. doi: 10.
1145/989863.989940
[2] N. Andrienko, G. Andrienko, and P. Gatalsky. Exploratory spatio-
temporal visualization: an analytical review. Journal of Visual Lan-
guages & Computing, 14(6):503–541, Dec. 2003. doi: 10.1016/s1045
-926x(03)00046-6
[3] A. Bojko. Eye Tracking the User Experience: A Practical Guide to
Research. 12 2013.
[4] S. Burigat and L. Chittaro. Visualizing the results of interactive queries
for geographic data on mobile devices. In Proceedings of the 2005
international workshop on Geographic information systems - GIS '05.
ACM Press, 2005. doi: 10.1145/1097064.1097103
[5] B. Carlo, B. Daniele, C. Federico, and G. Simone. A data qual-
ity methodology for heterogeneous data. International Journal of
Database Management Systems, 3(1):60–79, Feb. 2011. doi: 10.5121/
ijdms.2011.3105
[6] W. Chen, Z. Huang, F. Wu, M. Zhu, H. Guan, and R. Maciejewski.
VAUD: A visual analysis approach for exploring spatio-temporal urban
data. IEEE Transactions on Visualization and Computer Graphics,
24(9):2636–2648, Sept. 2018. doi: 10.1109/tvcg.2017.2758362
[7] H. Doraiswamy, E. T. Zacharatou, F. Miranda, M. Lage, A. Ailamaki,
C. T. Silva, and J. Freire. Interactive visual exploration of spatio-
temporal urban data sets using urbane. In Proceedings of the 2018
International Conference on Management of Data - SIGMOD '18.
ACM Press, 2018. doi: 10.1145/3183713.3193559
[8] G. Emberger, P. Pfaffenbichler, and L. Riedl. MARS meets ANIMAP:
Interlinking the model MARS with dynamic internet cartography. Jour-
nal of Maps, 6(1):240–249, Jan. 2010. doi: 10.4113/jom.2010.1079
[9] D. Fernondez-Prieto, C. Naranjo-Valero, J. T. Hernandez, and H. Hagen.
STRAD wheel: Web-based library for visualizing temporal data. IEEE
Computer Graphics and Applications, 37(2):99–105, Mar. 2017. doi:
10.1109/mcg.2017.27
[10] F. Gravetter and L. Wallnau. Essentials of Statistics for the Behavioral
Sciences. Cengage Learning, 2013.
[11] J. A. Guerra Gómez, J. C. Ortı́z Román, and J. G. Murillo Castillo.
Navio: a visualization widget for summarizing, exploring and navigat-
ing large multivariate datasets. 2018.
[12] S. Lavalle, E. Lesser, R. Shockley, M. S. Hopkins, and N. Kruschwitz.
Big data, analytics and the path from insights to value. MIT Sloan
Management Review, 52:21–32, 12 2011.
[13] T. Munzner. Visualization Analysis and Design. AK Peters Visualiza-
tion Series. CRC Press, 2015.
[14] J. R. Peña Bastidas, L. A. Guzmán Garcı́a, and J. A. Carrasco. Linking
a walking and a land use model to support expansion policies in cities:
A case study in bogotá. 12 2018.
[15] H. Senaratne, M. Mueller, M. Behrisch, F. Lalanne, J. Bustos-Jimenez,
J. Schneidewind, D. Keim, and T. Schreck. Urban mobility analysis
with mobile network data: A visual analytics approach. IEEE Trans-
actions on Intelligent Transportation Systems, 19(5):1537–1546, May
2018. doi: 10.1109/tits.2017.2727281
[16] H. Sequeira. Spatio-temporal visualizer: Online tool to visualize
spatiotemporal data with an interactive time-window slider, using d3.js,
leaflet.js and crossfilter. 07 2014.
[17] T. von Landesberger, S. Bremm, N. Andrienko, G. Andrienko, and
M. Tekusova. Visual analytics methods for categoric spatio-temporal
data. In 2012 IEEE Conference on Visual Analytics Science and Tech-
nology (VAST). IEEE, Oct. 2012. doi: 10.1109/vast.2012.6400553
[18] R. Y. Wang and D. M. Strong. Beyond accuracy: What data quality
means to data consumers. Journal of Management Information Systems,
12(4):5–33, Mar. 1996. doi: 10.1080/07421222.1996.11518099