Please enable JavaScript.
Coggle requires JavaScript to display documents.
Clustering (Consideration for Cluster Analysis (Separation of clusters,…
Clustering
- Consideration for Cluster Analysis
- Exclusive
- each object belongs to just one cluster
-
-
-
-
Hierarchical
- Multi level
- non-exclusive
- K --> not required
- Distance based
- Dandorgram
-
Density based
- Density function.
- Connectivity based
-
Partitioning
- Single level
- Exclusive
- K -> number of clusters
- distance based
-
-
- Application:
- Stand alone tool.
- Preprocessing step --> Handle outliers.
- Stand alone application:
- Biology
- Information retrieval
- Land use
- Marketing
- City- Planning
- Earth-quake studies
- Climate
- Economic Science
- Preprocessing tool::
- Summarization
- Compression
- Finding K-nearest Neighbors
-Outliers :explode:
- Quality (good): :star:
- Intra-class --> High similarity --> distance is minimized.
- Inter-class --> low similarity --> distance is maximized.
-
- Quality depends on:
- Similarity measure.
- Implementation.
- Ability to discover the hidden patterns.
-
-