Part 2. MM4XL Tools > 2. Analytical Tools > Cluster Analysis > How to run Cluster Analysis

Cluster Analysis

How to run Cluster Analysis

Tip:
It's easy to get up to speed with MM4XL in a matter of seconds:

  1. Select the MM4XL tool you want to work with from the toolbar.
  2. Click the Example button on the form. This opens an example sheet with data for the tool you are using.
  3. With your mouse, select the data sets as explained in the example sheet, then click the OK button to print the report.

From the MM4XL menu select Cluster Analysis and the window below displays.

Cluster Analysis Software for Marketing Segmentation

In the Input range select the range, using your mouse, where the input data is stored, and select an Output range where to print the resulting output, starting at the upper left cell. If your data set does not have row and/or column labels, uncheck the corresponding option in the frame.

The option Cluster data vertically changes the orientation of the Ward analysis. When unchecked, the analysis is done using input data by column, otherwise it is done using input data by row.

The Normalize input data (Standardization) option is checked by default. For more information about standardizing data see later in this chapter.

Clicking on the tab K-means or Wards method selects the respective clustering method. For the K-means method the Number of clusters is needed as input from the user. The frame Terminate Condition hosts two options for stopping the cluster algorithm from reiterating. Find exact solution stops when the best partitioning is reached and, default option, the user can stop the algorithm at the Best of n iterations. Each time the algorithm runs, items are assigned to a certain group and a new ratio (Inertia Between Group) / (Inertia Within Group) is computed. The highest ratio of all iterations is then chosen as the best partition.

Tip:
The order of data entry changes the partition. This is due to the random start seed the cluster analysis uses. Therefore, when working with the option Best on n iterations it is suggested to repeat the K-means several times and choose the partition with the highest Trace(B) / Trace(W) ratio.

The Dispersion chart is a quick visual aid that shows how items are clustered in groups. Print cluster numbers prints, in the first available column beside the input data set, a new column of numbers corresponding to the cluster group each item belongs to.

Cluster Analysis Software for Marketing Segmentation

The Wards method is very easy to run. The user can simply accept the default setting and click on the OK button. But one can also determine when and how to terminate the algorithm re-iteration. The first option, Automatically, stops when all items are grouped in one cluster only. Alternatively, the user can stop the tool as the desired Number of groups is reached. The third option stops the algorithm when the desired Inertia level is reached. For using the latter two options, simply check the corresponding option and type an integer value in the input box. If you make an incorrect entry, MM4XL warns you, do not be afraid to try.

Note:
Although the Find exact solution option produces the most accurate partition, it may take a *huge* amount of time in order to test all re-arrangements of the input data. 10 items measured on 4 variables may take up to 1 minute. Then time grows exponentially as new items are added.

Lifetime license:

Price: euro

Vote this tool
We proudly serve
Your vote
vote1 vote2 vote3 vote4 vote5
Email:
Gender:
M
F
Age:
Position:
Department:
Comment: