KB230547: How to create a cluster analysis report in MicroStrategy Enterprise Manager 9.x?

This technical note explains how to create a cluster analysis report in Strategy Enterprise Manager 9.x.
The out-of-the-box sample report is included in the Tutorial Project under \Public Objects\Reports\Strategy Platform Capabilities\Strategy Data Mining Services\Cluster Analysis. The segmentation is seen in a report and the result of the model is seen in the predictive view model in a graphical way as shown in the image below. The K-Means algorithm is used in the predictive metric.
 

 
User may want to use hierarchical clustering approaches. The following steps show how to create a hierarchical cluster report using R with the Tutorial Project.
 

• Create the independent metrics for Customer attribute, and the metric to show customer’s first name in a dendogram.
  First Name: Max(Customer@[First Name]) {~+}
  Age Range ID: Max([Age Range]@ID) {Customer+, ~+}
  Education ID: Max([Education Level]@ID) {Customer+, ~+}
  Gender ID: Max([Customer Gender]@ID) {Customer+, ~+}
  Housing Type ID: Max([Housing Type]@ID) {Customer+, ~+}
  Marital Status ID: Max([Marital Status]@ID) {Customer+, ~+}
   
• Create an R Script as follows:
  

  
  #Get the data
  if(exists("mstr.ExFlag")) { 
  #If this is executed by Strategy
   df <- data.frame(cbind(col0, col1, col2, col3, col4, col5))# data frame from the input variables   
   rownames(df) <- df[, 1]        # Input row name from the first column
   df2 <- df[ ,-1]                # Delete the first column.
   rownames(df2) <- rownames(df)  # Set the row name
  } else {                                                     
  #If this is NOT via a Strategy Report Execution
  df<- matrix(c(1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4), nrow=4, ncol=4)
      rownames(df)<-c("A","B","C","D")
      numofclusters="5"
  }
  #Modeling
  model <- dist(df2)        # distance matrix
  cluster <- hclust(model, method="complete")   #complete linkage 
  cluster.k <- cutree(cluster, k=numofclusters) # cut tree into the number of clusters  
  png("../Images/hc.png")   # the device to save a png image
  plot(cluster)             # draw dendogram
  dev.off()                 # close the device
  

  
  
• Open the script in a deployR tool and generate the metric expression by editing the parameters as shown in the image below: 
  

• Add the following line between the #Strategy_BEGIN and #Strategy_END lines in the script.
  

  
  #RVAR numofcluster -parameter StringParam9

  
   
  
• Create a metric with the expression at Step_3 and right-mouse click ‘RScript’ in the expression to open RScript Parameters. Input the number of clusters to be segmented in the value of StringParam9, as shown in the image below:
  

• Replace the input parameters with the metrics created in Step_1 as shown in the image below:
  

• Create a report with the following attributes and the metric. Add the out-of-the-box, ‘n-th Record Sample’ filter to the report.
  Customer
  Age Range
  Education Level
  Customer Gender
  Housing Type
  Marital Status
   
• The report with sample 20 records shows 5 clusters as specified in the parameter at Step_6.
  

• Create a Report Service document and specify the image outputted from R Script to see a dendrogram.