A SOLUTION TO A MULTIDIMENSIONAL DYNAMIC DATA ANALYSIS PROBLEM BY THE VISUALIZATION METHOD

D.D. Popov1, I.E. Milman1, V.V. Pilyugin1, A.A. Pasko2

1National Research Nuclear University MEPhI (Moscow Engineering Physics Institute), Moscow, Russian Federation

2National Centre for Computer Animation, Bournemouth University, Bournemouth, United Kingdom

dpopovmephi@gmail.com, igalush@gmail.com, VVPilyugin@mephi.ru, apasko@bournemouth.ac.uk

 

Contents

1. Introduction

2. Formulation of the problem

2.1. Description of the dynamic geometrical processes

2.2. Formal description of the problem solution process by the visualization method

2.3. Visualization of source data

3. Description of the solution algorithm

4. Description of the application program

4.1. Options of the application program

4.2. Examples of use of the application program

5. Conclusion

References

 

Abstract

The article describes a solution of a data analysis problem. Data to be analyzed represent changes in a given set of multidimensional objects with time. We propose to apply the visualization method to solve this problem. A formalization of the method is presented with a mathematical description of each stage of the source data visualization.

A developed interactive visualization application program is described. It is based on the models of the theoretical part of the article.

We emphasize the efficiency of the visualization method. It allows one to make a judgment on the formation of bunches or clusters of objects formalized in the form of n-tuples of real numbers and to find the objects seeking to be in a cluster or a bunch. Additionally, examples of use of the developed program for searching for invariants in changing the source data are provided.

 

Keywords: multidimensional analysis, dynamic data analysis, visual analysis, multidimensional visual analysis.

 

1. Introduction

 

There is an urgent problem of processing and analysis of multidimensional data in the modern world. For its solution there are developed many different methods and means, both automatic and interactive. Visual methods occupy a special place among data analysis methods.

However, a careful study of the literature devoted to the description of particular applications that use visual methods suggests that interactive systems working with multi-dimensional data are often valued less than systems that depict results of application of the Data Analysis methods. For example, there are such systems as the system of situational notification AdAware [1], the system of visual analysis in aircraft design problems [2], the visual analysis of textual information VxInsight, software package SAS Visual Analytics [3] designed for processing and analyzing large volumes of financial and economic information. All these systems are of an industrial nature are commercial, the systems provide the user with a great number of interfaces and data visualization capabilities. However, at the same time, all these systems are adjusted to the internal processing of these multidimensional data and presentation of these data to the user in a convenient way, without giving to the user either a possibility to work directly with the cloud data or to work with multivariate visual representation of the data [4, 7].

Theoretical generalizations of a solution of a problem of a source data analysis using the visualization method that was based on scientific data are considered in [5]. This method can be divided into the two following stages that in general can be repeated:

and

The original analysis algorithm of multidimensional geometric data based on this method is presented in [6]. A visualization application program based on this algorithm was developed. This application’s main feature is the possibility to work with multidimensional source data directly. The analyst purposeful manipulate directly with the source data and perform visual analysis of the results without making any original numerical processing of the source data. The paper shows that the application program can effectively solve a problem of multidimensional static source data analysis.

However, in practice we have to deal very often with multidimensional dynamic source data: monthly or annual organizations’ reports, elementary particles’ properties in different time periods, etc. These data contain information about the progress of these objects with time. The analyst wants to make a judgment about this progress, i.e., a judgment about the multidimensional dynamic source data. This article discusses developed mathematical models of dynamic multidimensional source data and the interactive application program that allows the analyst to analyze the data by the visualization method.

 

2. Formulation of the problem

 

Dynamic source data are values of some quantitative characteristics of these objects, which can change during the time. Each object can be denoted by an n‑tuple of real numbers in a fixed moment in time. We will consider the n-tuples of real numbers as points in a multidimensional Euclidean space  with a defined distance. Thus we assign a multidimensional dynamic data analysis problem by the visualization method with a geometric interpretation, i.e. the task of the analysis of change in the relative position of the points in the space .

Subsets of points representing clots and clusters can be allocated in this space, these subsets are described in [4]:

A cluster - a subset within a given point set, where the pairwise distance does not exceed the pre-defined d value and the distance between cluster points and other points is not less than the pre-defined d value.

A bunch - a subset of points with most distances between points not exceeding the preset d value.

In a particular way subsets may consist of a single point, in [4] the following points classification is given:

A remote (single) point - a point distant from all other points of the initial set for more than the preset d value.

Quasi-remote (quasi-single) point - a point that is not remote but at the same time is not included in a bunch or a cluster at the given grouping.

Isolation of clots and quasi-remote points performs by a man in the process of solving analysis problem.

Note that remote and quasi-remote points are special cases of clusters and clots respectively.

The values of the coordinates of points can change over time. Points can form clots and clusters or join them over time and vice versa.

The main purpose of this work is the solution of the analysis problem of changes of the mutual arrangement of the assigned set of the points in space  by the visualization method. In achieving this purpose, a solution of a multidimensional dynamic data analysis problem by the visualization method will be found.

The solving of the problem can be divided into the following stages:

·       Making a mathematical description of the objects analyzed.

·       Development of an algorithm for a solution of a multidimensional dynamic data analysis problem by the visualization method.

·       Writing an application to solve the analysis problem.

Let us introduce the following notions that will be used to describe the subsequent material.

A Geometric Process is a set of points in space, coordinates of which are time-dependent.

We call a spatial process the variable spatial scene depending on the time. In other words, a spatial process is a dynamic spatial scene. Detailed description of the scene will be given below (paragraph Mapping).

 

2.1. Description of the dynamic geometrical processes

 

The initial object of the analysis is a set of n-dimensional points . Coordinates of the points are given in several time instants: .

       A point has n coordinates:

the distance for each pair of point equals:

We have the discrete geometrical process in the beginning: , it consists of discrete processes  – dynamic n-dimensional points, which are given by  – their dynamic coordinates.

The process  is a chronological geometrical description of states of the items involved. However, the items change continuously. We use an interpolation to describe the continually changing items.

Process  were interpolated using piecewise linear interpolation technique. In this case, the desired interpolation function for time-dependent coordinates of points  is binomial  in closed interval , . For a value ,   is given from the equation:

       The result of the interpolation  is a geometrical continuous process. Then  is a temporal section,  belongs to the domain of .

        can be discretize using temporal sections. The discrete geometrical process is the collection of temporal sections for selected . So  is  temporal sections of the  process.

 

2.2. Formal description of the problem solution process by the visualization method

 

       We use the visualization method [5] to solve this problem. The method is based on the sequential solution (generally multiple) of the two tasks presented below:

 

Macintosh:Users:Popov_Dmitry:Desktop:ÓÍË:Íàóêà:Ñòàòüÿ:Áåç çàãîëîâêà.png

Fig. 1. The visualization method

 

The analyst specifies visualization parameters and obtains static or animated images (i.e chronological frames) until he/she is able to make his/her own judgment on changes in the relative position of points with time. Thus a solution of a multidimensional dynamic data analysis problem by the visualization method is interactive.

A visualization application program should provide an interactive user interface. The analyst should have an opportunity to influence a spatial scene and obtained images. This makes solving of the problem much more efficient. Thus a solution of a multidimensional dynamic data analysis problem by the visualization method is iterative and interactive.

The visualization process of the source data will be reviewed below.

 

2.3. Visualization of source data

 

Visualization of source data is a sequence of the following steps, which is called visualization pipeline:

 

Fig. 2. Source data visualization

 

Sourcing of dynamic data

 

The discrete geometric process  is specified in this step. This process be visualized later on.

 

Filtering

 

In this step, the source data are preprocessed.

The initial discrete process  is interpolated using piecewise linear interpolation technique as described earlier. The result of filtering is an obtained continuous process .

 

Mapping

 

The Analyst selects a 3-dimensional subspace of the n-dimensional space that will be used to create a continuous spatial processes. are numbers of basis vectors of the original n-dimensional space. They form the basis of the subspace.

The initial set of points is projected onto the selected subspace.

  is the set of projections, .

After that, the analyst selects the radius of spheres  that are associated with the obtained 3-dimensional points, their color  and the radius of cylinders  that connect the spheres with each other. The distance between the connected spheres is less than the assigned d. associating with points the distance between which is less than the assigned d.

Spatial scene  corresponds to the obtained continuous geometric process , which is a result of interpolation of the source discrete geometric process .  is the description of the scene geometry and  is the description of the optical parameters of the scene. The scene is a continuous spatial process:

.

Then at each fixed ,  will correspond to , . Let us define.

where  is a sphere of radius  centered at the point  is a cylinder of radius that connects two spheres, .

where   is a color of the spheres,  are colors of the cylinders.

 is a color and opacity of the -th cylinder that connects spheres associated with points . The color is defined in RGB, the opacity is defined in RGB as well:

where  means that the cylinder is completely transparent, % – means that the cylinder is completely opaque.

Spatial discrete process space  can be created from this continuous process as described above for the geometric continuous process .

For example, this option may be necessary if the analyst is interested only in the key moments which show a clot or a cluster formation.

 

Rendering

 

The projection image  is the result of a scene’s  rendering. A scene is a process:  . We have the scene and, therefore, the projection image for each time t. Let us enter one more process .

where  is a set of visualization attributes.

Visualization attributes are a camera , light, physical characteristics of the scene environment, size of an obtained image, etc. They can be either static or dynamic.

By camera we mean the point of view which fix the scene  .  where  is the camera position,  is the focus and  is the camera angle.

The sequence of frames  is the result of discrete spatial process’s  rendering. Frames  can be used as key frames for the construction of the animation.

 

3. Description of the solution algorithm

 

The algorithm for solving the problem consists of the following steps:

In case if the obtained projection images are not enough, and the analyst cannot make judgment about interesting in his opinion data, the algorithm provides returns to the scene setting stage and to the stage of setting visualization attributes.

The algorithm described above is shown below in Figure 3.

 

Algoritm.png

Fig. 3. Solution algorithm

 

4. Description of the application program

 

We developed an interactive visualization application program to solve the problem of the analysis of changes in the relative position of points. The application program is based on the algorithm shown above.

 

4.1. Options of the application program

 

The application program gives the analyst the following options:

The developed application program provides user-friendly and interactive visual interface that allows the analyst to manipulate the rendered space scene.

We used 3ds Max® application program and its internal object-oriented programming language MAXScript.

 

4.2. Examples of use of the application program

 

The program was tested using the source data containing monthly reports of credit institutions.

 

Macintosh:Users:Popov_Dmitry:Desktop:ÓÈÐ:Ïðèìåð 2:Ïðèìåð ðàáîòû 2.jpg

Fig. 4. Images from the "sections gallery" obtained for the credit institutions’ data.

 

The moment when the distant point joint the clot was discovered as a result of the images (Fig. 4) analysis. The judgment that has been made is that the point of the subset became the point of another subset during the time .

After analyzing animation shown in Figure 5, we can make an interesting conclusion.

 

Macintosh:Users:Popov_Dmitry:Downloads:ïðîáà (ïðåîáðàçîâàí).gif

Fig. 5. The result of dynamic spatial process rendering

All sectors associated with the initial n-dimensional points are approximately located in one plane.

This observation enables us to make the following judgment. The equation  is right for the projections of points' coordinates  on the selected subspace .  are constants.

Source data was calculated by the method of least squares, and the following equation of approximating plane was carried out:

In the examples above, numbers correspond to the coordinates which values reflect the following financial indicators of credit institutions:

Note that a similar approach, based on an approximation of the planes in the space of principal components used in [8,9].

 

5. Conclusion

 

In this paper, we have:

The interactive user interface allows a user to find the time when points form clusters and clumps and vice versa. The developed application program provides user-friendly interactive visual interface that allows the analyst to manipulate the rendered space scene.

We can say that the developed application program performs dynamic animated visualization of multidimensional source data. This application program automatically creates a description of the key spatial scenes, which are animated then.

In addition, the user is given an opportunity to cut out an interesting in his opinion part of the process. He can control the speed of the process’ flow in the resulting animation. These options are available interactive.

The nature of the dynamic source data may require another type of the data interpolation, thus we plan to make it possible to choose an another interpolation technique in the next version of the application program. Another direction of the research is geometric and spatial processes of operator variable.

 

References

 

1.     Livnat Y., Agutter J., Moon S., Foresti S. Visual correlation for situational awareness. IEEE Symposium on Information Visualization. pp. 95-102, 2005.

2.     Mavris D., Pinon O., Fullmer D.Jr. Systems design and modeling: A visual analytics approach. 27th Congress of International Council of the Aeronautical Sciences ICAS, 2010.

3.     SAS the power to know. URL: http://www.sas.com/en_us/home.html. [Access date: 26 1 2016].

4.     Maslennikov O.P., Milman I.E., Safiullin A.E., Bondarev A.E., Nizametdinov Sh.U., Pilyugin V.V. Razrabotka sistemy interaktivnogo vizualnogo analiza mnogomernykh dannykh [Development of a system for analyzing multidimensional data]. Scientific visualization. V.6, no. 4, p. 30-49, 2014. (in Russian)

5.     Pilyugin V., Malikova E., Pasko A., Adzhiev V. Nauchnaja vizualizacija kak metod analiza nauchnyh dannyh [Scientific visualization as method of scientific data analysis]. Scientific Visualization. V. 4, no. 4, pp. 56-70, 2012 (in Russian)

6.     Milman I.E., Pakhomov A.P., Pilyugin V.V., Pisarchik E.E., Stepanov A.A., Beketnova Yu.M., Denisenko A.S., Fomin Ya.A. Data analysis of credit organizations by means of interactive visual analysis of multidimensional data. Scientific Visualization. 2015. V. 7, no. 1, pp. 45 – 64

7.     Maslennikov O.P., Milman I.E., Safiullin A.E., Bondarev A.E., Nizametdinov Sh.U., Pilyugin V.V. Interaktivny vizualny analiz mnogomernykh dannykh [Interactive visual analysis of multidimensional data]// GraphiKon'2014: 24th International conference on computer graphics and vision: Rostov-on-Don, the SFU Academy of architecture and arts, Conference materials. - p. 51-54 (in Russian)

8.     Bondarev A.E, Galaktionov V.A. Parametric Optimizing Analysis of Unsteady Structures and Visualization of Multidimensional Data. International Journal of Modeling, Simulation and Scientific Computing. 2013. V.04. N supp01. 13 p. DOI 10.1142/S1793962313410043.

9.     Bondarev A.E. Analiz mnogomernyh dannyh v zadachah vychislitel'noj gazovoj dinamiki [Multidimensional data analysis in cfd problems]. Scientific visualization. Vol. 6. No. 5. Pp. 61-68. 2014.