Scribd
Upload
10 views

4.Data Mining - Pattern Mining in Multilevel, Multidimensional Space, Rare and Negative Patterns

The document discusses methods for mining multi-level and multi-dimensional association data, focusing on quantitative association rules and the identification of rare and negative patterns. It outlines techniques for mining associations at various abstraction levels, including the use of concept hierarchies and support-confidence frameworks. Additionally, it addresses redundancy in association rules and provides approaches for mining multidimensional and quantitative association rules, as well as methods for discovering rare patterns.

Uploaded by

umakanthn
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
10 views

4.Data Mining - Pattern Mining in Multilevel, Multidimensional Space, Rare and Negative Patterns

The document discusses methods for mining multi-level and multi-dimensional association data, focusing on quantitative association rules and the identification of rare and negative patterns. It outlines techniques for mining associations at various abstraction levels, including the use of concept hierarchies and support-confidence frameworks. Additionally, it addresses redundancy in association rules and provides approaches for mining multidimensional and quantitative association rules, as well as methods for discovering rare patterns.

Uploaded by

umakanthn
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Pattern Mining in Multilevel,

Multidimensional Space,
Rare and Negative patterns

By,
Umakanth N
Session Outcomes

 Mine the multi-level and multi-dimensional


association data
 Mine Quantitative association rules; and mine
data with rare and negative patterns

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 2
Road map on Pattern Mining
Research

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 3
Mining Multilevel
Associations
 Developing effective methods for mining patterns
at multiple abstraction levels. Eg., Stationary
related transaction
 Level 1: Stationary
 Level 2: Pen, Pencil
 Level 3: Parker Pen, Ink Pen, Ball-Point Pen
 Association rules generated from mining data at
multiple abstraction levels are called multiple-
level or multilevel association rules
 Can be mined efficiently by using concept
hierarchies under a support-confidence framework
(By calculating frequent itemsets for all abstraction
levels using Pattern
February 10, 2025
any Mining
algorithm)
in Multilevel, Multidimensional Space, Rare and Negativ 4
Mining Multilevel
Associations
 Methods to find frequent itemsets for each
abstraction levels
 Using uniform minimum support for all levels:
 Same min_support threshold is used for all abstraction
levels

 Drawback: May mislead in finding meaningful


associations
February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 5
Mining Multilevel
Associations
 Methods to find frequent itemsets for each
abstraction levels
 Using reduced minimum support at lower levels:

 Using item or group-based minimum support:


 Set up user-specific, item, or group-based minimal
support thresholds

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 6
Checking redundancy among
multilevel association rules

 rule can be considered as redundant if its


support and confidence are close to their
“expected” values

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 7
Mining Multidimensional
Associations
 Single dimensional or intra-dimensional
association rule
 Buys(X, “digital camera”) => Buys(X, “HP printer”)
 Multi dimensional association rule
 Inter-dimensional association rules:
 Age (X, “20 – 29”) ^ Income(X, “20k to 35k”) =>
Buys(X, “Laptop”)
 Hybrid-dimensional association rules
 Age (X, “20 – 29”) ^ Buys(X, “Laptop”) => Buys(X,
“Pendrive”)

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 8
Mining Multidimensional
Associations
 Techniques for mining Multidimensional
Association rules:
 Static discretization of quantitative attributes
 Replacing the numeric values with discretized range of
values
 Quantitative association rules
 Quantitative attributes are discretized or clustered into
“bins” based on the data distribution

 For multi-dimensional data, frequent


predicate-set {Eg., (age, income,
buys) } needs to be calculated instead of
frequent item set.
February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 9
Mining Quantitative
Association Rules
 Quantitative attributes can be discretized into
multiple intervals and then treat them as nominal
data in association mining.
 But it leads to large number of association rules,
which may not be useful.
 Methods to discover novel associations:
 Data cube method,
 Clustering-based method,
 Statistical analysis method (To disclose exceptional
behavior)

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 10
Data cube method
 Multidimensional data may be used to construct
a data cube:

 It can store aggregate values in multi-


dimensional space; which is essential for
calculating support-confidence values for multi-
dimensional attributes
February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 11
Clustering based
method
 k-means (or) density-based clustering algorithm
can be used to calculate the clusters for the
quantitative data
 Top-down approach:
 Find clusters in 1-D that satisfies min_support
 Combine diff. clusters to generate 2-D cluster
 …..
 Bottom-up approach

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 12
Mining Rare and Negative
Patterns
 An infrequent (or rare) pattern is a pattern with a
frequency support that is below (or far below) a
user-specified minimum support threshold

February 10, 2025 Pattern Mining in Multilevel, Multidimensional Space, Rare and Negativ 13
? ?
. . ?
ie s
e r
Q u

You might also like