This is the continuation of the previous article. For better understanding it is recommended to look through the earlier blog posts.
Data visualization types and objectives
Nowadays, the amount of data being produced and stored increases rapidly. Due to the rapid development and low cost of data transmission and storing tools the amount of stored data is steadily reaching enormous values. For instance, according to the data provided by the Berkley University the whole data being gathered all over the world increments by 1 million terabytes annually.
Naturally, in order to process all this data and process information ‘usefully’ it is necessary to develop more complicated methods and, in some cases, new fields of study.
Moreover, from the point of view of processing speed more powerful calculation capacity is required.
Modern development of cloud computing allows to access, in the short period of time, almost unlimited resources for data processing and representation, and, in case of shortage, to broaden and deepen the needed data. It is also necessary to mention one more factor that should always be taken into consideration while creating data processing systems. This factor is determined by the human brain’s specific work behavior. By certain reasons, our brain is able to easily process a large amount of data presented as graphs, pictures and flowchart. But nevertheless, data presented in numeric or textual format is extremely hard to be processed by our brain, especially when we are talking about large datasets.
To sum up, all these factors increase the necessity of creating new methods and tools of data visualization of high quality.
Data visualization is a combination of methods of data representation which is characterized by a specific format. This format is always easy to analyze and process. Data visualization often increases the user perception rate and facilitates the data management.
There are several types of visualization:
Quantitative data representation in diagram form: pie charts and linear diagrams, histograms and spectrograms, tables and graphs etc.
Conceptual visualization that allows developing complicated concepts, plans and ideas with the help of concept maps, Gantt charts, shortest pathway graphs etc.
Strategic visualization that transforms different data concerning an enterprise activity: productivity diagrams, life-cycle diagrams and work breakdown structure diagrams
Metaphoric visualization, which means structured information represented as pyramids, binary trees and data maps (e.g. metro subway scheme)
Combined visualization, allowing combination of several complicated graphs in one single scheme (e.g. weather forecast graph)
By their goals visualization techniques may be classified as presentation or research visualization. Presentation visualization is aimed to show data to certain audience (e.g. during scientific research, lecture or analytical overview). The main goal of research visualization is to provide data in such a form that it could be later analyzed and processed in order to find new possible solutions.
There are also hybrid presentation/research data visualization forms. In this case, besides providing a user with visualized data, we give him a possibility to explore provided information by means of special interactive tools allowing, for instance, a user to impose restrictions on data.
Due to the number of dimensions used in model construction, data visualization methods may be divided into two groups:
Data representation in 1, 2 or 3 dimensions
Data representation in 4, 5 or more dimensions
But still, this division is conditional on the human ability to percept multidimensional objects.
Generally, well-known data representation methods belong to the first group. If we are talking about Data Mining, almost all the methods and tools commonly used in this field of study are taken from the first group.
According to the number of dimensions used in data representation, they may be classified by 3 categories:
One dimensional, or so-called, 1-D
Two dimensional, 2-D
Three dimensional or projection, 3-D
Two and three dimensional representation is the most natural way of data visualization for human beings. While using these methods of representation the most demonstrative sets of tools are:
object distribution by class and cluster structure;
topological characteristics;
existence of trends;
information about mutual data colocation;
existence of other dependencies specific to the explored dataset
In case of existence of more than three dimensions in the dataset more sophisticated data visualization techniques are required. It is possible either to use multidimensional data representation methods or to decrease the representation level to 1, 2 or 3 dimensions. In order to decrease the number of dimensions and to visualize data at the same time the Kohonen self-organizing maps are commonly used.
It still should be mentioned that methods allowing perception of data in four or even more dimensions have been developed long ago. These methods are, for instance, parallel coordinates, Chernoff faces method, petal diagram method.
All of the methods mentioned above are efficient in application and require users’ high level competence in a particular field of study and profound understanding of the working principles of these visualization methods. But in our opinion in the sight of renewable energy sources monitoring all the methods turn out to be too complicated. In this case a profound preliminary analysis of datasets and their division by parameters suitable for representation in two or three dimensions is considered to be more efficient.
Before applying a certain method it is necessary to analyze if all the data is to be visualized or it is possible to define the most valuable portion. It is also necessary to choose the method that satisfies the entire requirements list specific to the dataset.
Among 2 or 3 dimensional tools of visualization the most popular are linear graphs, linear diagrams, bar graphs, pie charts, and sector/vector diagrams.
From the point of view of the studied topic one of the most interesting directions is state-spatial representation which is considered as a separate field of visualization. In most cases with the help of these tools separate regions on the map are defined and highlighted with different colors according to the analysis indicator.
Data visualization subsystem is an essential part of sophisticated intellectual analysis systems, especially of those ones aimed to process big amount of data.
To be continued.