Now shipping ClustanGraphics 8 - check here  We now offer a polythetic divisive clustering procedure that finds the key variables to discriminate between cluster subsets most efficiently, as shown right - first key variable in mammals' milk is Lactose, second is Protein.  Check out Cluster Keys here. We have also published a new ClustanGraphics Primer, Fourth Edition, 2006, details here
Ready to cluster ... If you're interested in cluster analysis then you have reached the right site.  If not, hit the back button now, and we bid you a fair farewell.  Because Clustan is the home of cluster analysis .  We try to cluster well.  We try to make it easy and visual.  We know our solutions are flexible and powerful.  And because we are focused, you won't need any extra baggage - just put your data on a text file, in a spreadsheet, or copy it to the clipboard, and you're ready to cluster. Our software runs under Windows 95, 98, 2000, NT, ME and XP.  We also offer a Sun Solaris version and run on Macs via a PC emulator.  It is self-contained, indefinitely licensed, integrated with your Windows applications, and above all, dedicated to cluster analysis - no "excess baggage" requirement. ClustanGraphics 8 is our most popular clustering product yet.  First released in 1997, our 2005 version is now much faster and more versatile than ever.  This is because we work with our users, to provide the flexibility you need and demand.  Where other programs fail to deliver the features you require, Clustan CAN... Versatility is our strength.  ClustanGraphics 8 offers hierarchical agglomerative cluster analysis, hierarchical divisive cluster analysis, k-means analysis, focalpoint clustering, outlier analysis and proximity analysis.  You can also construct a cluster model from a hierarchical cluster analysis and then classify any number of new cases by reference to it.  We have been working on cluster analysis methods for over 30 years, often breaking new ground with our research publications, so you can be confident that our algorithms are at the top of the range. Flexibility is our goal. Your data can be simple or complex, complete or partially complete, small or massive.  You can compute general similarities on mixed data, euclidean distances on ordinal data, or binary similarities on dichotomous data.  You can read your own proximities if you computed them or observed them externally, and your proximity matrix can be incomplete. Simplicity is our promise.  For a straightforward cluster analysis using our recommended options, simply deploy the standard defaults of the Clustan Wizard - three clicks and you're clustered.  The wizard reads your data from an Excel spreadsheet , completes a hierarchical agglomerative cluster analysis, identifies the best cut, displays the resulting dendrogram shaded at the largest significant partition, and constructs a cluster membership table, ... all with just three clicks of the cursor. For an overview of the comprehensive features ClustanGraphics offers, click here.  If you have questions, check our FAQs page, browse our site map, ask at our guestbook, or simply read on .... Data Mining is a particular strength.  We can cluster 200,000 cases (or more) hierarchically, and run k-means on a million cases using a PC.  Mixed data types and missing values frequently occur in databases and social surveys, so you don't have to "warp" your data to fit ClustanGraphics.  And it's reassuring to know that when you hit a really large cluster analysis application, KEY FEATURE k-Means Tree ClustanGraphics can scale up to it.  If you're looking for powerful analytical tools, great flexibility, excellent graphics, with point-and-click ease of use, then look no further.  Because Clustan CAN... k-Means Analysis now includes a choice of criterion functions, outlier deletion, differential weights, treatment for mixed data types and provision for missing values .  Our k-means algorithm has an exact relocation test that is guaranteed to converge, unlike other k-means programs which can oscillate between sub-optimal solutions - see naïve k-means.  Read our critical appraisal, because "k-means" means Clustan CAN ... Mixed Data Types can be used in hierarchical cluster analysis, k-means and proximity analysis.  You can specify whether your variables are binary, nominal, ordinal or continuous, and whether they are to be transformed or differentially weighted.  Click here for details, and bookmark Clustan CAN... Missing Values are allowed in data or proximities.  You can read an incomplete data matrix and compute proximities from it.  You can also read an incomplete proximity matrix.  Then you can cluster with missing values using any of 11 hierarchical clustering methods.  k-means analysis, outlier and proximity analysis, cluster models, truncation and cluster profiling are all possible with incomplete data or proximities.  When it comes to handling messy, complex data, remember that Clustan CAN ... Multidimensional Scaling ClustanMDS implements Bell Labs' MDSCAL and KYST, which we have translated, simplified, and optimized for Windows - three clicks and you're configured.  You can read or compute a proximity matrix and then display the relationships between your cases or clusters on a scatterplot, to aid graphical visualization of your results.  Details here. FocalPoint Clustering is our two-stage k-means procedure with random trials, sponsored by McKinsey and Co.  It was presented at three conferences last year.  Details here. Whisky Classified is a market segmentation of single malt whiskies (right).  It has been presented at CSNA, IFCS, BCS and RSS, and at whisky festivals and seminars.  Our scientific presentation is followed by a tasting of single malt whiskies now sponsored by over 50 Scottish distillers.  A book "Whisky Classified: Choosing Single Malts by Flavour" has been published by Pavilion Books, London.  Second edition, March 2006, and in 5 language editions, available here. What Is Cluster Analysis + Hierarchical Clustering + Reading Proximities + Data Types + Computing Proximities + Multidimensional Scaling + Nearest Neighbours + Navigate Tree + Reorder Tree + Cluster Keys + Cluster Validation + Clustering Variables + Cluster Profiles + k-Means Analysis + k-Means Tree + FocalPoint Clustering + Cluster Exemplars + Missing Values + Large Data Clustering + Classify Cases + Cluster Scatterplots + ClustanGraphics Primer + Customer Relationship Management + Banking + Data Mining + Network Analysis Clustan CAN... so there's really no reason to hesitate.  Let's get clustering! Clustan - A Class Act © 1998 Clustan Ltd .