• Classification: classifying data into classes using Fuzzy k-means algorithm
• Allocation: allocating new data into existing classes (created using FuzME classification)

Classsification

Input/Output
Data-File: Select a file containing your data. See the definition page for the description of the required file format.
Output: Select an output directory by choosing it from the directory tree window. The program will produce all the output in this directory. Define an output file name in the "Output Filename" box. If the "Write seperate file for each class" box is checked the program will produce seperate file for each group/class containing the membership values of the data.

FuzME output is a file created by FuzME which will be used in allocation problem (allocating new data to classes).

Fuzzy Clustering

Choose the metric distance, number of classes, fuzzy exponent and the algorithm for the fuzzy clustering. See the definition page for more information.
 You can perform multiple runs of FuzME using different values of fuzzy exponent (phi) from "Start" to "End" with "Step" incerement.

For fuzzy k-means (other than FkM with extragrades) you can calculate fuzzy discriminat analysis. See fuzzy discriminat analysis for more details.

Advanced user: For fuzzy k-means with extragrades you can specify the average extragrade membership to find an optimum alfa. FuzME uses average extragrade value of 1/(no_class+1) as a default to find the value of alfa.

Jackknife

If the "Perform Jackknife" box is checked the program will do a Jackknife analysis on the defined number of classes. The method involves sampling certain amount of data from the whole dataset and clustering it with Fuzzy k-means for a certain number of runs. As a general guide, the minimum number of data to be left out of the analysis should be less than the square-root of the no. of data. (Efron, B. and Tibshirani, R.J. 1993. An Introduction to the Bootstrap.Chapman & Hall, NY.) Under the 'Number of data' box, enter the no. data & the % of data need to be sampled is automatically calculated. The centroid file contains the centrois of each classes from each number of runs.

Allocation

You need to specify the FuzME file to allocate new data into existing classes. Note. the number of attributes and the order of the data should be the same with the data used in classification.

More details are given in Definition page

FuzME
Download FuzME
ACPA