DATA MINING Desktop Survival Guide by Graham Williams
Desktop Survival Project Home Data Mining Introduction Getting Started Data Loading Data Exploring Data Interactive Graphics Test Descriptive Data Mining Predictive Data Mining Evaluation and Deployment Data Cleaning Handling Missing Data Transforming Data Data Reduction Deployment Troubleshooting Issues Moving into R Beyond Rattle R Getting Help Data Graphics in R Understanding Data Preparing Data Descriptive and Predictive Analytics Issues Evaluating Models Reporting Topics in Data Mining Fraud Analysis Archetype Analysis Text Mining Survival Analysis Algorithms Bagging Bayes Classifier Cluster Analysis Conditional Trees Hierarchical Clustering K-Nearest Neighbours Linear Models Support Vector Machines Open Products AlphaMiner Borgelt Data Mining Suite KNime R Rattle Weka Closed Products C4.5 Clementine Equbits Foresight GhostMiner InductionEngine ODM Enterprise Miner Statistica Data Miner TreeNet Virtual Predict Appendices Installing Rattle Bibliography Index Preface Goals Organisation Features Audience Typographical Conventions A Note on Languages Currency Acknowledgements

Hierarchical Clustering

Agglomerative clustering is used to build a hierarchical cluster. A complete hierarchical cluster is built on the click of the Execute button. You do not need to re-execute on changing the Number of Clusters. This simply needs to obtain the relevant information from the fully built hclust. But users will automatically go to re-execute after changing this (because this is how everything else in the interface works). An alternative is being considered to make it more obvious not to re-execute.

Once a cluster has been built, have a look at the dendrogram to visually get an idea of the ``natural'' number of clusters, and then set the number appropriately, then have a look at the stats and the plot.

Having a large number of entities to build the hierarchical cluster from is not sensible. Certainly 20,000 entities would not produce anything interpretable. Even perhaps 10,000 is beyond what is usefully interpretable. But there are techniques that help interpret. See http://www.ncbi.nlm.nih.gov/projects/geo/gds/analyze/analyze.cgi?datadir=UCorrelationUPGMA&ID=GDS3254&myType=0

The amap package includes standard hierarchical clustering with a choice of distances like Eulidean and Spearman, and a parallel implementation.

Support further development through the purchase of the PDF version of the book.
The PDF version is a formatted comprehensive draft book (with over 800 pages).
Brought to you by Togaware. This page generated: Saturday, 16 January 2010