UET
Since 2004
ĐẠI HỌC CÔNG NGHỆ, ĐHQGHN
VNU-University of Engineering and Technology
INT3137-Visual Data Analytics
Lecture 8: Visual Analytics
for Understanding Temporal Distributions and
Variations
Duc-Trong Le (PhD)
Hanoi, 04/2025
Lesson Plan
● Motivating Example
● Specifics of Temporal Phenomena and Data
● Transformations of Temporal Data
● Temporal Filtering
● Analysing Temporal Data with VAs
○ Events
○ Univariate Time Series
○ Time Series of Complex States
Duc-Trong Le INT3137 - Visual Data Analytics 2
Remind: VA for Understanding Relationships
Movie distribution by age, gender for each genre
Duc-Trong Le INT3137 - Visual Data Analytics 3
Motivating Example - Vastopolis Epidemic (1)
Temporal frequency
of disease-related
microblogs, i.e.,
events, using
histograms
Duc-Trong Le INT3137 - Visual Data Analytics 4
Motivating Example - Vastopolis Epidemic (2)
Duc-Trong Le INT3137 - Visual Data Analytics 5
Motivating Example - Vastopolis Epidemic (3)
Temporal frequency with coloured segments
of different event types
Duc-Trong Le INT3137 - Visual Data Analytics 6
Motivating Example - Cherry Blossoms (1)
Metadata records of geolocated photos of
cherry blossoms taken on the territory of the
North America in the years from 2007 till 2016
Duc-Trong Le INT3137 - Visual Data Analytics 7
Motivating Example - Cherry Blossoms (2)
34 clusters that include at least 50
elementary events (the sizes of the
clusters obtained range from 57 to 711
members) in four regions
Duc-Trong Le INT3137 - Visual Data Analytics 8
Motivating Example - Cherry Blossoms (3)
Duc-Trong Le INT3137 - Visual Data Analytics 9
Motivating Example - Cherry Blossoms (4)
Statistical summary
Duc-Trong Le INT3137 - Visual Data Analytics 10
Motivating Example - Cherry Blossoms (5)
Duc-Trong Le INT3137 - Visual Data Analytics 11
Specifics of Temporal Phenomena and Data (1)
● Events:
a) are things, actions, etc., that happen at some times or exist during a limited
time period, e.g., microblog posts or taking photos of cherry blossoms
b) can also be a succession of time steps in the life of an evolving phenomenon
whose state (properties and/or structure) changes over time
● Time series: is a chronological sequence of states of such a phenomenon
Duc-Trong Le INT3137 - Visual Data Analytics 12
Specifics of Temporal Phenomena and Data (2)
Phenomena and events often are dependent on temporal cycles, e.g., weekly, monthly,
seasonal, etc., which are domain-specific.
Duc-Trong Le INT3137 - Visual Data Analytics 13
Specifics of Temporal Phenomena and Data (3)
Most temporal phenomena have the property of temporal dependence, in which values or events
that are close in time are usually more similar or more related than those that are distant in time
Duc-Trong Le INT3137 - Visual Data Analytics 14
Specifics of Temporal Phenomena and Data (4)
The time range of the temporal dependence is often modified by temporal cycles
Duc-Trong Le INT3137 - Visual Data Analytics 15
Temporal Phenomena and Data Analysis
● Temporal data analysis is to discover the existence of temporal dependence and
cycles
● Methodology:
○ Statistic-based Method
○ Visualisation-based Method
● Some issues:
○ wrong recorded values
○ wrong time references,
○ and missing data
Duc-Trong Le INT3137 - Visual Data Analytics 16
Transformations of Temporal Data (1)
34 clusters that include at least 50
elementary events (the sizes of the
clusters obtained range from 57 to 711
members) in four regions
Integration transformation of multiple
elementary events
Integration is usually applied to
temporal clusters of events, e.g.,
regions, tree species, etc.
Duc-Trong Le INT3137 - Visual Data Analytics 17
Transformations of Temporal Data (2)
Aggregation transformation of a set of events by time intervals into time series of event
counts
Aggregation is to group events or multiple time series by time intervals
Duc-Trong Le INT3137 - Visual Data Analytics 18
Transformations of Temporal Data - Example (1)
People took more photos of
cherry blossoms at the daytime
than in the night and, generally,
more events happened on the
weekend than on the weekdays
Duc-Trong Le INT3137 - Visual Data Analytics 19
Transformations of Temporal Data - Example (2)
Smoothing or normalisation techniques could be applied to facilitate the comparison
between regions
Duc-Trong Le INT3137 - Visual Data Analytics 20
Possible Transformations of Temporal Data
Duc-Trong Le INT3137 - Visual Data Analytics 21
Temporal Filtering
● Temporal filtering approaches:
○ Selecting a continuous time interval
within the time range of the data
○ Filter data according to the positions of
their time references within one or more
time cycles
● Some tricks :
○ Simultaneous temporal filtering
○ Different filters to different datasets
○ Make use of appropriate visualisations
Duc-Trong Le INT3137 - Visual Data Analytics 22
Analysing Temporal Data with VAs - Event
Three kinds of analysis tasks referring to event sequences
● In a single event sequence, observe patterns of event repetition.
● Having a large number of event sequences, find common patterns of
event arrangement.
● Having a set of standard event sequences with the same relative
ordering, detect temporal anomalies
Duc-Trong Le INT3137 - Visual Data Analytics 23
Analysing Temporal Data with VAs - Event
Repeated sub-sequences in a single sequence
Distance functions may be utilised to match sub-sequences
Duc-Trong Le INT3137 - Visual Data Analytics 24
Analysing Temporal Data with VAs - Event
Common patterns of event arrangement in multiple sequences
Alignment and Aggregation operations are utilised to find common patterns
Duc-Trong Le INT3137 - Visual Data Analytics 25
Analysing Temporal Data with VAs - Event
Common patterns of event arrangement in multiple sequences
Duc-Trong Le INT3137 - Visual Data Analytics 26
Analysing Temporal Data with VAs - Event
Anomalies in standard event sequences
The train stops are positioned
along the vertical axis according
to their distances from each
other. The horizontal axis
represents time. The train
journeys are represented by
diagonal lines running from top
left to bottom right (Paris – Lyon)
and bottom left to top right (Lyon
– Paris) respectively
Duc-Trong Le INT3137 - Visual Data Analytics 27
Analysing Temporal Data with VAs - Event
Anomalies in standard event sequences
The data need to be represented so that
anomalies (that is, deviations from the
ideal procedure) are easy to spot.
Duc-Trong Le INT3137 - Visual Data Analytics 28
Analysing Temporal Data with VAs - Univariate Time Series
The heatmap technique may be better suitable for long time series and
for multiple time series that need to be compared
Duc-Trong Le INT3137 - Visual Data Analytics 29
Analysing Temporal Data with VAs - Univariate Time Series
Different variation patterns may exist at different temporal scales
Duc-Trong Le INT3137 - Visual Data Analytics 30
Analysing Temporal Data with VAs - Univariate Time Series
Motif discovery techniques may also be helpful for finding anomalies
Duc-Trong Le INT3137 - Visual Data Analytics 31
Analysing Temporal Data with VAs - Univariate Time Series
The data describe the variation of the power demand of a facility during a year with a hourly
temporal resolution
Duc-Trong Le INT3137 - Visual Data Analytics 32
Analysing Temporal Data with VAs - Time Series of Complex States
Two approaches are applicable:
a) Clustering of similar states and representation of the temporal
distribution of the clusters using temporal displays;
Duc-Trong Le INT3137 - Visual Data Analytics 35
Analysing Temporal Data with VAs - Time Series of Complex States
Two approaches are applicable:
a) Clustering of similar states and representation of the temporal
distribution of the clusters using temporal displays;
Duc-Trong Le INT3137 - Visual Data Analytics 36
Analysing Temporal Data with VAs - Time Series of Complex States
Duc-Trong Le INT3137 - Visual Data Analytics 37
Analysing Temporal Data with VAs - Time Series of Complex States
Two approaches are applicable:
b) Embedding (or projection): positioning of the states on a plane
according to their similarities and representation of the state progression
as a trajectory.
Duc-Trong Le INT3137 - Visual Data Analytics 38
Analysing Temporal Data with VAs - Time Series of Complex States
Duc-Trong Le INT3137 - Visual Data Analytics 39
Summary
● Motivating Example
● Specifics of Temporal Phenomena and Data
● Transformations of Temporal Data
● Temporal Filtering
● Analysing Temporal Data with VAs
○ Events
○ Univariate Time Series
○ Time Series of Complex States
Duc-Trong Le INT3137 - Visual Data Analytics 40
UET
Since 2004
ĐẠI HỌC CÔNG NGHỆ, ĐHQGHN
VNU-University of Engineering and Technology
Thank you
Email me
[email protected]
