Lecture 7 Flashcards
Time Oriented Data
Time is an inherent data dimension that is central to the task of revealing trends and patterns in the data.
The important point to note while working with time oriented data in order to find pattern and trends are:
Chose a visual representation that fits the data characteristics
Parametrize the visual representation accordingly in order to be able to detect patterns hidden in the data.
General Aspects of Time
Scale: Ordinal vs. discrete vs. continuous
Scope: Point-based vs. interval-based
Arrangement: Linear vs. cyclic
Viewpoint: Ordered vs. branching vs. multiple perspective
Scale
The first aspect of time is ‘Scale’.
Divided into three parts:
Ordinal – Relative order relation (before, during, after)
Discrete – Temporal distances are considered
Based on smallest possible time (minutes, seconds)
Most commonly used Continuous – A possible mapping to real numbers Between two points in time, another point in time exists
Scope
Second aspect to consider about time is scope
It can be of two types:
Point based
Have a temporal extant equal to 0
Can be seen in analogy to discrete Euclidean points in space
No information is given about the region between two points in time
Example May 1, 2014 00:00:00
Interval based
Relates to subsection of time having temporal extent greater than 0
Example [May 1, 2014 00:00:00, May 1, 2014 23:59:59]
Arrangement
Third design aspect of time is arrangement
Categorized into two:
Linear
We mostly consider time as proceeding linearly from past to future
Each time value has a unique predecessor and successor
Cyclic
Time domain is composed of a set of recurring time values.
Any time value A is succeeded and preceded at the same time by any other time value B
E.g. Winter comes before summer but winter also succeeds summer
Viewpoint
Fourth subdivision
Can be of the following types:
Ordered – Consider things that happens one after the other
Totally ordered – Only one thing can happen at a time.
Partially ordered – Overlapping events are allowed
Branching – Multiple strands of time branch out and allows description and comparison of alternative scenarios, only one path through time will actually happen, e.g. Project planning
Multiple Perspective – Facilitates simultaneous views like eye witnesses report of an incident
Hierarchical Organization of Time
Hierarchical organization of time and concrete time elements is determined based on
Granularity and calendars: None vs. Single vs. Multiple
Time Primitives: Instant vs. Interval vs. Span
Determinacy: Determinate vs. Indeterminate
Granularity and Calendars
It can be thought of as a human made abstraction of time in order to make it easier to deal with in everyday life.
Examples like minutes, hours, days, weeks, etc.
If a granularity and calendar system is supported by time model, we characterize it as multiple granularity
If every time value is given in terms of milliseconds it is single granularity
If none of the abstractions are supported then its none
Time Primitives
Intermediary layer between data elements and time domain
It can be of three types:
Instant – A point in time
Interval – An interval of time, have fixed start point and end point
Span – Durations without a fixed position e.g. 6 days
The first two are anchored or absolute primitive
Span is unanchored or relative primitive.
Determinacy
Uncertainty is one aspect which needs to be considered while working with time oriented data
If there is no exact information about time specification or if time primitives are converted from one granularity to another, uncertainties are introduced.
E.g. Time when the earth was formed, one or two days ago
If we say an activity started on May 1, 2018, it is certain if granularity is ‘day’ but when the granularity changes to ‘hours’ we are uncertain as at which hour the activity started.
