9  Filtering flowdata

This chapter concerns the filtering procedures for flowdata in order to reduce the graphic complexity of the flow map - the so-called spaghetti effect.

Filtering procedures are also important to select useful information: links and/or nodes to obtain an interesting flowmap.

Without filtering, the flowmap patterns reveals a spaghetti-effect. See example

EXAMPLE: RICardo international trade flows like a dish of spaghetti.

The selected information shown is described in percent on the bottom of the right panel. Here, all the information is represented :100% of the links,100% of the nodes and 100% of the total information).

Filtering can be performed in Arabesque on all variables describing the nodes and/or links.

9.1 Available filtering procedures

9.1.1 Filtering possibilities

Filtering is available on numerical, temporal and categorical variables.

9.1.1.1 Numerical variables

Numerical filtering applies to quantitative (absolute, continuous or pseudo-continuous) dat is done either visually with a brush on an interactive histogram or bar chart. It is also possible to indicate a threshold.

9.1.1.2 Temporal variables

The filtering possibilities of the temporal matrices are currently only available for links. They concern the choice of a date or a period, in a numerical or graphical form (slider). Three date formats are available (string, hours, timestamp).

9.1.1.3 Categorical variables

Categorical filtering is applied to nominal qualitative data available in node and/or link datasets. Three possibilities are offered, graphically and numerically:

Selection of multiple variables: the user chooses one or more of the available categorical variables (e.g., prefectures and sub-prefectures if a variable specifying the administrative profile of French cities is available) or the type of links (e.g., import or export for trade data);

Selection of a single variable: the user chooses a single variable (for the nodes: for example, the flows from and to France for mobility data; the import type flows for trade data);

Selection by deletion of a single variable: the user chooses one or more variables relating to the nodes and links to be deleted. If the deletion selection concerns nodes, all the corresponding links (for example, the Ile-de-France region for an analysis at the national level) are deleted.

9.1.2 Implementing filtering

The Arabesque filtering possibilities depend on the type of variable/procedure of filtering, as well as the corresponding graph for visual implementation.

Three main possibilities are proposed

– (1) Visual filtering using modular selection area (brush) applied on the histogram of the distribution of links for a given variable.

– (2) Numerical filtering by entering a minimum (and/or maximum) threshold value.

– (3) Visual filtering for categorizes data, by selecting one or more values from a drop-down list