This workflow can be found on the KNIME Workflow Public Server under
This KNIME workflow focuses on identifying classes of telecommunication customers that churn using K-Means. As in all exploratory data mining, it is unknown beforehand what number of clusters will be appropriate, therefore the workflow allows the user to specify different numbers of clusters for K-means to calculate. Each set of clusters has its’ cluster centers and variances combined and presented both as a report and automatically into a spreedsheet so that clusters can be interpreted.
This workflow uses QuickForms and flow variables to provide runtime flexibility, a metanode that can be reused for combining the cluster information, report generation as well as integration with Microsoft Office products.
The churn data is available as a free download from Ian Pardoe via www.iainpardoe.com/teaching/dsc433/data/Churn.xls