Visualizing Patterns of Flow and Usage in Wireless Networks

This information visualization research depicts the connections and transitions of users, the streams and patterns of flow and the internal structure of wireless networks. The following images and videos display the data of the network campus of the University of Coimbra regarding the academic year of 2010-2011. The research is documented in the Master’s dissertation entitled Visualizing Patterns of Flow and Usage in Wireless Networks.
 

 

The Circuit of Transitions

 
The Circuit of Transitions is a visualization inspired by the radial convergence layout to represent the network’s structure and the user transitions. The graph emphasizes the common user routes and the frequently accessed points in wireless networks.
 

The nodes are arc blocks positioned along a circumference accordingly to the transitional proximity. The network’s access points (APs) are grouped in clusters accordingly to their reachability in the temporal interval of 60 seconds. The height of each node is proportional to the number of grouped APs.
 
The edges are uni-directional Bézier curves, thinner on bends and larger on extremities. Each edge connects two nodes (the source node and the target node) and groups all transitions occurred between them. The transitions within the same AP or cluster are drawn outside of the main circle, while the edges that connect distinct nodes are drawn inside the circumference. The width of the edge’s extremity is proportional to the amount of grouped transitions. The edge inherits the color of the node of origin to represent the transitional direction. Thus, the length of each node block is proportional to the sum of its incoming and outgoing transitions. The height of the bend point of an edge is proportional to the average transition time value (mapped accordingly to the log10 scale). The incoming and outgoing transitions of a node are organized by width.
 
In the absence of spatial information about the localizations of APs, the implemented clustering algorithm groups nodes by proximity, using the average transition time associated between node pairs. Also, we order node positions by relative spatial proximity based on the average transition time value. To promote clarity and efficiency, we provide a filtering technique that removes rarely accessed APs and rarely occurred transitions, below a given threshold.
 
The novel interactive technique allows users to expand or collapse any cluster, allowing them to control the level of detail of the visualization. When a cluster is collapsed, the transitions of all its nodes are represented by the cluster. When expanded, one can observe the incoming and outgoing transitions of the individual nodes represented at the cluster’s position. Pointing with the mouse cursor over a node shows its ID number and its technical name. The user may turn on all node IDs at once, expand top five biggest clusters and filter the displayed data using the respective controllers included in the interface.
 

The Circuit of Transitions represents the global movement of users within a wireless network, portraying the flow and the patterns of user transitions. The node clustering and ordering algorithms allow the grouping and sorting of the APs by relative proximity, providing visual clarity to the network’s structure and displaying additional information. The graphic system permits the identification of the more common (thick edges) and the less common (thin edges) routes, the short (low curves) and the long transition (high curves) paths, the origin and the destination point of each transition, the most (bigger nodes) and the least (smaller nodes) accessed APs and clusters of APs. The video at the top of this page reveals how the network’s structure and connectivity changes over time.
 
 

The Matrix of Connections

 
The Matrix of Connections is a visualization inspired by bi-dimensional matrices to represent the user connections in wireless networks. The graph displays the patterns of user connectivity with the density of connections.
 

The list of users is shown along the vertical axis, sorted in descending order regarding their total connection times. The connections of a user are red traces mapped on the timeline (the horizontal axis) accordingly to the “start” and the “stop” events. The application allows the visualization of separate and joint connections. The joint connections represent the total connection times of the network’s users, being displayed in a horizontal bar chart. The temporal grid ticks appear behind the red lines of connections, and the respec- tive labels are situated above and below the graph’s area. In the separate connections’ graph, the ticks represent real days, weeks and months. In the joint connections’ graph, the ticks divide the timeline by days and weeks only, since week is a standard and common time unit with the biggest invariable length.
 
The mouse cursor offers the exploration of the information. Pointing the cursor on the connections of a user highlights the user line and shows the user’s ID number on both sides and the technical name on the left side of the graph. This feature works for both separate and joint connections’ graphs. Additionally, the same action reveals the user’s total connection time on the right side of a bar, when the joint connections are displayed. A vertical line is added to the actual cursor’s position to facilitate the localization of the extremities of the connections on the timeline, to compare different connections situated at the same level, and to show the temporal values accordingly to the aiming position. While exploring the separate connections’ graph, the mouse pointer indicates the desired date and time at the top and bottom grid label’s section. While interacting with the joint connections’ graph, the cursor shows the values of total time, counting from the starting point to the current position of the pointer. These values are also in the grid label’s zone and follow the pointer’s position as well. The user can navigate through the graph by performing a dragging gesture or using the scroll bars. Both types of navigation are synchronized and can be used as alternatives. The user may switch on all user IDs at once, enable the horizontal grid lines and scale the graph using the interface controllers.
 

The Matrix of Connections displays the user connections within a network. The overall and particular connectivity is revealed by duration, density and periodicity of accesses. A graphic of total connection times can also be visualized. The order of the users shows the hierarchy in information, while the timeline localizes the temporal intervals each user spent online. The implemented zooming and the navigation techniques provide the exploration of the information. The model depicts the patterns of connectivity and the specific temporal periods regarding the visual density of connections. The common timeframes of access to the network are shown with the vertical repetition of the red traces, located on the grid.
 
 

The Panels of Flow Shapes

 
The Panels of Flow Shapes is a visualization inspired by small multiples representation. This model reveals the patterns of usage in wireless networks due to various data aggregation combinations, as well as the patterns of human behavior in those networks.
 

The graph presents the data aggregated in three levels of temporal granularity. The possible granularities are month, week, day, hour and minute. There can be visualized two data categories (connections and transitions), two data types (times and amounts), four time variables (minimum, average, maximum and total) and two amount variables (users and records).
 
The visualization consists of a bi-dimensional matrix with multiple cells. One axis of the matrix holds the 1st data aggregation level, another axis represents the 2nd aggregation level. The graph’s columns and rows are labeled accordingly to the aggregation levels. One axis of a cell shows the 3rd aggregation level, and another axis of a cell holds data values, mapped accordingly to the log10 scale. The cells have equal sizes and are uniformly spaced. Both axes have small white-colored grid ticks based on the currently represented values. Multiple variables are shown with layered area graphs with different colors. The visualization has a horizontal view (reading along the columns of the matrix) and a vertical view (reading along the rows of the matrix).
 

The interaction with the visualization is performed by the use of the mouse cursor. While interacting with the panels, the user can highlight any desired cell with the mouse pointer. The color of the area graphs becomes more saturated, the labels appear along the grid ticks, and the data points are shown at the mapped positions. When the user points the mouse cursor on a data point, the respective exact value appears above the cursor in the numeric format. If the user performs a left click on a cell, the cell stays highlighted, allowing a comparison of the desired panels. The user may enable the grid lines, change the reading view and deselect all cells at once using the respective interface controllers.
 

The Panels of Flow Shapes demonstrate peaks, lows and rhythms in the connectivity and transitionality of users on wireless networks. The amounts of connections/transitions can be compared with the quantities of users, and four values of connection/transition time (minimum, average, maximum and total) can be compared between themselves. The patterns of usage of the network are then revealed by choosing three time granularity levels, being shown with two impositions.
 
 

The Map of the Wireless Network

 
The Map of the Wireless Network is a visualization inspired by laws of Physics, depicting a wireless network as a system with an organic movement. The graph displays the flow of users and the physical structure of the network. The human behavior is presented by the user particles.
 

The network’s APs (colored nodes) are grouped in clusters (white nodes with colored stroke) accordingly to their reachability in the temporal interval of 60 seconds. The size of each node is proportional to the amount of incoming and outgoing user transitions. The clusters are affected by a repulsion force that separates them visually. The child particles of a cluster are affected by three forces: the attraction force that pulls them to the center of the cluster; the first repulsion force that separates them visually; and the second repulsion force that pushes them away from the border of the father cluster.
 
The graph’s edges are uni-directional mechanical springs shown with lines. Each edge connects the pair of nodes and groups all transitions occurred between them. The thickness of an edge is proportional to the amount of grouped transitions. The edge inherits the color of the node of origin to represent the transitional direction. The length of an edge is proportional to the average transition time. Therefore, all nodes are distributed in space accordingly to the transitional proximity.
 
The users are small gray circles that move around the nodes they are actually connected to. Each user has a network path that presents the route between visited nodes in the given temporal period.
 

The visualization has a variety of interactive functions. Pointing the cursor on a cluster makes appear its ID number (in the bold type weight) above the particle and the amount of children below the mouse pointer. The incoming and outgoing edges of that cluster become visible and highlighted with a bigger opacity. The user can expand a cluster double-clicking on it. The contour and the background of the expanded cluster become semi-transparent, while the children nodes become opaque. The name, the ID (in the regular type weight) and the incoming and outgoing transitions of a child can be visualized by pointing the cursor on the child node’s circle. When a cluster is expanded, the user particles connected to it relocate to the respective children APs. Additionally, a cluster can be affixed in the space with a left-click on its shape. This functionality prevents the node from moving under the force of springs and allows an easier exploration of the cluster’s children and edges. When the mouse cursor points on a user circle, the user’s ID number appears above the particle and the performed network path is shown. That path consists of tracing semi-transparent thin black lines between visited pairs of APs/clusters. The user may switch all node IDs on, make visible the transitional paths, detach all nodes at once and stop all moving particle using the respective interface controllers.
 

The Map of the Wireless Network shows the structure of the network and the flow of its users. The system’s behaviors are inspired by the laws of Physics, providing an organic movement to the graph’s elements. All elements are affected by balanced forces with different effects and strengths. The clustering technique creates visual groups of APs. The user move around the nodes they are connected to. The pairs of nodes are distributed in space due to the mechanical force of the connecting edge with distance proportional to the average transition time. The common APs and clusters are highlighted with bigger sizes, and the common transitional routes are emphasized with bigger thicknesses. The video below demonstrates the temporal flow of users on the wireless network of the University of Coimbra.
 

 
 

In Proceedings

• R. Kamolov, P. Machado, and P. Cruz, “Visualizing the flow of users on a wireless network,” in Special Interest Group on Computer Graphics and Interactive Techniques Conference, SIGGRAPH ’13, Anaheim, CA, USA, July 21-25, 2013, Poster Proceedings, 2013, p. 114.

  • Bibtex
  @inproceedings{kmc13a,
  author = {Ruslan Kamolov and Penousal Machado and Pedro Cruz},
  title = {Visualizing the flow of users on a wireless network},
  booktitle = {Special Interest Group on Computer Graphics and Interactive Techniques Conference, SIGGRAPH ’13, Anaheim, CA, USA, July 21-25, 2013, Poster Proceedings},
  year = {2013},
  pages = {114},
  ee = {http://doi.acm.org/10.1145/2503385.2503510},
  bibsource = {DBLP, http://dblp.uni-trier.de},
  publisher = {ACM},
  isbn = {978-1-4503-2342-0}
  }
  • |
  • PDF