During the design of various processes, including
those associated with the presence of a large number of internal relationships,
the volume of heterogeneous data is formed, i.e. a set of heterogeneous data
characterizing the entire process under study and its individual elements (Fig.
1). The interpretation of such data is aimed at identifying implicit patterns
and contradictions in heterogeneous descriptions of processes [1].
![[G2019]001](pics/image001.jpg)
Fig. 1 Standard representation of
educational process parameters
A large amount of input data and a limitation of
available resources (computational, time, human) often become problems of
practical analysis of heterogeneous data during designing and controlling
results of various processes. The search for new analysis tools is conditioned
by the need to analyze a large number of relationships among elements of the
general process [2]. Traditional means have several disadvantages, which
include high qualification requirements to experts involved in the analysis and
the need for interdisciplinary interaction among them. Besides, to improve the achieved
results it is necessary to eliminate internal contradictions in the studied
data, the source of which is errors made at the process planning stage. The subjective
nature of such contradictions complicates their detection based on pre-formulated
criteria.
The amount of data and a number of additional
information influence the data interpretation process. Some information is not
directly related to the current problem. At the stage of preliminary analysis
of the initial data or during the problem solving process the analyst gets a
lot of additional information (intermediate solutions, verification data,
classification and clustering results, etc.). In addition, the process of
researching and interpreting data which requires significant time resources is
associated with the need to study new data Ö replacing or complementing the old
ones [3].
History of data acquisition, methods of their
collection, the presence of errors, partial lack of data, etc. can have a significant
impact on the research objective achievement. Metadata, i.e. second-order data,
from the point of view of interpretation, play the role of an additional source
of information for the researcher and are a part of heterogeneous data. Thus,
the peculiarity of studying heterogeneous data at the user's disposal when solving
a planning problem is the need to compare many heterogeneous descriptions of
individual elements of the process, to take into account intermediate versions
of solutions and to update interpreted data during analysis. As an advanced way
to overcome these difficulties, it is proposed to use data visualization tools capable
of organizing an interactive communication between a user and initial data [4].
Interpretation of heterogeneous data is a process
of obtaining new information and it is characterized by the amount of resources
needed for this. This condition is formed by the requirements for the data
interpretation tools. Interactive visualization used to represent and interpret
the source data makes it possible to use the potential of visual perception to
compensate for these difficulties. An important feature of visual perception is
the ability to compare several events (options), as well as the credibility of
the received information, which is the basis for making necessary decisions.
Visual image of data can be examined in various
ways and, therefore, can give the user answers to various questions. In the
simplest case, visualization provides confirmation or refutation of the already
formulated answer to the question and does not require additional understanding
of the results interpretation. This allows us to highlight the illustrative
role of visualization, which has a pronounced practical orientation, for
example, demonstration of the results of solving analysis problems obtained by
other means. This type of visualization is characterized by the development in
the direction of automated model construction, as well as the search for visual
representation forms reducing interpretation time.
The paper presents methods for more complex use of
visual models based on the direct interaction between the researcher and the visualization
tool. The opportunity to change the state of the visual model while it is used corresponds
to the construction of a controlled data model and creates a wide range of
advantages, which include the targeted use of the model state to monitor the
control action results [5].
The analyzed process is an ordered set of elements
O={O1..On}, where Oi is an
elementary process defined by a set of unique properties [6]. Meanwhile, properties
of the element Oi are divided into two subsets, based on
their functional differences:
Oi = {Pi}=
{Inx, Outy} ,
where the subset {Inx} is
process requirements, so-called incoming connections of Oi element,
and the subset {Outy} is results of the process, so-called outgoing
connections of Oi element. Connections of Oi
element determined in this way are its main characteristics which ensure
interaction with other elements and an achievement of the goal of the main
process.
In accordance with these definitions, the goal of
the main process can be represented as a composition of outgoing connections,
characterized by some parameters, for example, the number of outgoing
connections (results) and some weighting factors. Thus, the potency of set O={O1..On}
is defined by the predetermined goal G={Out} and the available
resources.
The formal goal of solving the planning problem is
formulated as following: to find the set O that provides targeted or best
results and meets all boundary requirements:
O={On : n = min, G=max}.
In the simplest case, when properties of elements Oi
are specified, the goal of the solution is to order the set O which ensures
the fulfillment of the boundary requirements. In the opposite case, the
solution to the problem is a new set O* = {In*, Out*}.
Human reasoning abilities are diverse and
insufficiently studied [7], so, the following are highlighted for data
visualization:
Õ
Adaptability, i.e. readiness to include new information in the
system of existing concepts and their categories. A useful consequence of this
property is the ability to perceive rapidly changing visual information,
including observing scenes that are complex sets of objects and events, as well
as identifying interactions, including new ones.
Õ
Attention selectiveness, which excludes details from the
perceptual data set that are not in the focus of attention. This optimizes
activity in conditions of limited resources (for example, time).
Õ
A two-stage perception mechanism, which, interacting with
accumulated experience, creates a flexible process of the operational comprehension
of a visual image.
The need to use perceptual abilities,
characterized by empirical principles, sets to a developer a task of
determination of the set of user characteristics that are purposefully involved
in the data visual image interpretation [8]. Selection can be based on a
preliminary assessment of available resources, which include:
Õ Characteristics
of a potential user, whose involvement in the process of data interpretation
does not create prerequisites for increasing its duration.
Õ Computing
resources, which allow obtaining visual images of data, if they meet the
requirement of interactivity with them.
Õ Time
resources that determine speed of construction and interpretation of visual
data images.
Õ Additional
requirements arising from the statement of the research task, including: the userÒs
prior knowledge, his qualifications, probable features of perception, etc.
The set of the userÒs characteristics involved in
the cognitive interpretation of the visual image of the studied data can be
determined by generalization of existing schemes of the visualization process [9].
Based on the obtained set, three groups of visualization tools are
distinguished, which differ in functional purpose and in methods of practical
implementation:
Õ Observation.
It is obtaining visual information (perception of colors, space, movements,
allocation of groups, recognition of forms, signs).
Õ Search.
Identifying relevant objects and processes (spatial thinking, prior awareness,
motivation) in the initial visual information.
Õ Formulation.
Formulating hypothesis of the answer to the research question (experience of
using visual analytics, ability to study and apply new language systems),
formalization of new information.
Development of computer visualization technologies
and their continuous complication create difficulties in interaction between a
user and the developed visualization tools [10]. This circumstance becomes
critical in a situation where visualization tools provide cooperative
participation of a group of researchers or are a way of exchanging information
between experts with different levels of training or area of specialization.
There is a need to choose between training users to use new visualization tools
or involving existing visual communication skills in interpreting data.
Based on the formulated requirements for
visualization tools, which are necessary to solve the planning problem, a
system for interactive presentation of data included in the description of an
arbitrary educational program has been developed. Three-dimensional visual
model is proposed as a visualization tool designed to solve the problem. It
forms a visual image of information objects included in the source data (Fig.
2). An information object is an element of an educational program (an academic
discipline, a course, program section). Each such object is an array of data
(name, course, duration, content, incoming requirements, planned results),
including variables of different types.
Software has been developed that allows acquiring
an interpretable visual image with the use of Autodesk 3ds Max package visualizers.
The algorithm for constructing a visual image is implemented using Maxscript language;
therefore, it is portable and can be easily adapted to new visualization
technical capabilities. Visualization interface only partially uses Autodesk
3ds Max environment and can be adapted to a particular userÒs needs.
![[G2019]006](pics/image002.jpg)
Fig. 2 Tool for interactive representation
of data, included in the description of an arbitrary educational program.
Reasonable interactive management system
creates conditions for setting new research questions and receiving answers
quickly, accelerating achievement of the analysis goal. Consequently,
interactive features of the visualization tool determine sequence and logic of
a researcher's reasoning.
To reduce the training period required to
become familiar with the new interpretation tool, it is proposed to use a
representation metaphor based on traditional methods of visualizing tabular
data (charts, graphs). A cylindrical coordinate system is defined in 3D space
of the visualization tool, that allows each point in space to be matched with
three values: training time, load, result. The scale of measurement units along
time and result axes can be arbitrary; the load is measured as percent of the
maximum possible. The opportunity to compare objects is realized through the
use of color coding, which rules can be changed in accordance with perceptional
characteristics of a particular user.
The radial direction of the time axis
allows visualizing data of educational programs of any duration (bachelor's
degree, specialistÒs program and master's degree). Program steps which correspond
to the given time intervals (years, semesters) are divided by concentric
circular elements used to represent the accumulated results. Each concentric
element is a reference scale (0-100%) with a common starting point. The
proposed structure provides presentation of an increasing number of learning
results without worsening general perception of data.
The information object provides to an
observer an opportunity to interpret visual attributes as values of the
corresponding parameters: color is an identification attribute, dimensions are
load ones, position corresponds to the training period. Information about
incoming requirements and planned learning results is presented as links
between information objects. Attributes of such connections are their direction
and quantity, which correspond to the source data. In accordance with
characteristics of the subject area, incoming (requirements) and outgoing
(results) communications have the same type, i.e. they can be interpreted as
created or developed competencies.
For visual representation of accumulated
learning results, simultaneous use of two expressive means is proposed: a
rating scale of effectiveness and visual scaling of links. In the first case,
it becomes possible to efficiently use the three-dimensional space of the
visual model, in the second one Ö the assessment of training results in the
educational program can occur while interpreting two-dimensional visualization.
![[G2019]009](pics/image003.jpg)
Fig. 3 Visualization of
accumulated results.
A method for visualizing intermediate and
general results in the form of result summing profiles was developed. A profile
has a form of a histogram representing accumulated results as a percentage of
planned values. To simplify the image of the studied data, elements of the
result presentation can be temporarily excluded from the visual model.
Visual objects, which present to a user
data of the achieved learning results, perform one of the main functions, which
consists in searching for options of individual element characteristics satisfying
the planning goal. Thus, the use of visual representation elements as a system
of interactive control of the state of the visualization tool is proposed. The
interface of interaction with the visualization tool created in this case
becomes the basis for cognitive interpretation of the source data images taking
into account individual characteristics of a user's thinking.
Dynamic correction of planned results by
detection and elimination of contradictions made at the planning stage can be
deemed a promising option for employing user interaction with the developed
visualization tool.
The quickest search occurs for contradictions of several types:
Õ
Chronological discrepancy between incoming requirements of the
information object and the results achieved (Fig. 4). Contradiction arises if
the input of the information object requires data that will be received later.
It is visualized as a line of accumulated results going in the opposite
direction.
Fig.
4 Chronological discrepancy.
Õ
Lack of input data: incoming requirements of the information
object are not provided with results, i.e. the requested data are missing. It
is detected as an open input of an information object (Fig. 5).
Õ
Unreasonable use of existing resources. It is a result of
interpretation of result lines that are not presented at the output of the overall
process.
![[G2019]002](pics/image005.jpg)
Fig.
5 Lack of input data
Õ
Degradation of results. The proposed visualization tool takes into
account possibility of reducing the level of the achieved result over the
course of time. The results present in the final profile that do not take into
account the level of their degradation are considered a contradiction.
Õ
Duplication of results. Resources are used for repeatedly
obtaining similar results. The visual image of such a contradiction is
diverging lines of cumulative results.
Application of the developed visualization tool has
made it possible to obtain the assessment of advantages of the proposed approach in comparison with traditional methods of
interpretation and verification of heterogeneous data contained in documents
regulating educational programs.
Table 1.
Comparison of duration of data research stages.
|
Comparison parameter
|
Traditional
approach
|
Visual
analysis
|
|
Training
|
Previous experience
|
Less than 5 min
|
|
Forming general idea
|
15-20 min
|
Less than 2 min
|
|
Searching contradictions
|
up to 30 min
|
1-5 min
|
|
Changing goal
|
Impossible
|
Interactive
|
|
Managing data
|
Impossible
|
10-20 min
|
The proposed tool for visual analytics of
educational environment can be supplemented with experience maintainability
when already formed versions of educational trajectories and corresponding real
results are saved. Opportunities of advanced planning for periods of varying
duration are expanded by analyzing factors that influence deviation of real
values of achieving professional competencies from those planned in educational
program design.
The use of visualization tools as means of
operational data research in planning tasks is an example of visual analysis of
heterogeneous data. The obtained advantages include high speed of visual
perception, which is necessary for simultaneous comparison of large amount of
disparate facts, as well as the opportunity to interpret data not only in
numerical form.
The technique of using visualization tools allows
achieving the research goal without interpreting numerical values. Thus, an
approach to solving problems of this type solely based on visualization tools
is proposed. The visualization tool has a low dependence on specifics of the problem
subject area. Thus, the experience of using the proposed means of
interpretation can be involved in solving other problems. Low requirements for usersÒ
preliminary preparedness make it possible to use visualization tools in absence
of special training for solving data interpretation problems which have an
interdisciplinary nature.
This work was supported by Russian Science
Foundation, project 18-11-00215.
1. Guo H. [et al.]. A Case Study Using
Visualization Interaction Logs and Insight Metrics to Understand How Analysts Arrive
at Insights // IEEE Transactions on Visualization and Computer Graphics, 2016,
Vol. 22, Issue 1, pp. 51Ö60. doi: 10.1109/TVCG.2015.2467613
2. Zakharova A.A., Vekhter E.V., Shklyar
A.V., Zavyalov D.A. Visual Detection of Internal Patterns in the Empirical Data
// // A. Kravets et al. (Eds.): CIT&DS 2017, Communications in Computer and
Information Science, Vol. 754, pp. 215-230. doi: 10.1007/978-3-319-65551-2_16
3.
Manakov D., Averbukh V. Verification of Visualization // Scientific
Visualization, 2016, Vol. 8, No. 1, pp. 58Ö94.
4. Blascheck T. [et al.]. VA2: A Visual
Analytics Approach for Evaluating Visual Analytics Applications // IEEE
Transactions on Visualization and Computer Graphics, 2016, Vol. 22, Issue 1, pp.
61Ö70. doi: 10.1109/TVCG.2015.2467871
5. Zakharova A.A., Vekhter E.V., Shklyar A.V.
Methods of Solving Problems of Data Analysis Using Analytical Visual Models //
Scientific Visualization, 2017, Vol. 9, No. 4, pp. 78Ö88. doi: 10.26583/sv.9.4.08
6. Zakharova A.A., Shklyar A.V., Rizen Y.S.
Measurable Features of Visualization Tasks // Scientific Visualization, 2016,
Vol. 8, No. 1, pp. 95Ö107.
7. Biederman I. Recognition-by-Components: A
Theory of Human Image Understanding // Psychological Review, 1987, Vol. 94, No.
2, pp. 115Ö147.
8. Chen H. [et al.]. Uncertainty-Aware
Multidimensional Ensemble Data Visualization and Exploration // IEEE
Transactions on Visualization and Computer Graphics, 2015, Vol. 21, Issue 9, pp.
1072Ö1086. doi: 10.1109/TVCG.2015.2410278
9. Van Wijk J.J. The Value of Visualization
// Proceedings of VIS 05. IEEE Visualization, 2005, doi: 10.1109/VISUAL.2005.1532781
10. Chen C. Top 10 Unsolved Information
Visualization Problems // IEEE Computer Graphics and Applications, 2005, Vol.
25 Issue 4, pp. 12Ö16. doi: 10.1109/MCG.2005.91
