Classification and clusters

Содержание и основные понятия

Classification
Using examples with known labels and choosing one of the specified classes.
Clusterization
Doesn't know the names in advance: she's looking for groups of similar objects, and the person is checking their meaning.
Grade 11
PIF · Baseline
Data analysis
01
Two performances
Classification
In terms of length and number of participants, identify a known type of school event: a lecture or a "workshop".
Clusterization
By the same tokens, find compact groups without finished names. The clusters received may not match the usual types of events.
02
Map of two features
XY
Duration x number of participants
<!--alterna-source:0JrQsNGA0YLQsCDRiNC60L7Qu9GM0L3Ri9GFINGB0L7QsdGL0YLQuNC5-->Map of school events
Horizontally, the length, vertically the number of participants.
Longitude, minutes
Number of participants
03
Boundaries and mistakes
The new object is classified as a standard class, for example, for the nearest marked neighbors. The object near the border is less certain. Errors of different classes may have different prices, so overall accuracy is not sufficient.
Verification
Classification
Clusterization
What is known in advance?
Class names and selected examples.
Just signs and a measure of resemblance.
How do I check it out?
Postponed sample, error matrix.
Group sustainability and meaningful meaning.
Main risk
Answer leak, class distortion.
Random or pointless groups.