This Application performs cluster analysis in multi-dimension space. The Application can use as input the output structure from the “t2ed_output.mat" produced by Transformation to Equivalent Dimensions Application therefore all analyses are performed within the Equivalent Dimension phase (parameter) space.

REFERENCES Code Repository

CATEGORY Data Processing Applications

KEYWORDS Data clustering, Production-seismicity interaction, Statistical analysis

CITATION Please acknowledge use of this application in your work: IS-EPOS. (2019). Cluster Analysis [Web application]. Retrieved from https://tcs.ah-epos.eu/  

Step by Step   

After the User adds the Application into his/her personal workspace, the following window appears on the screen (Figure 1):

Figure 1. Input data and parameters selection for “Cluster Analysis” Application.

The User is now requested to fill the fields shown in Figure 1 (from top to bottom):

• Input Data (T2ED output data) - The User may click on "change input" button in order to use a data file among the ones that are already uploaded or created within his/her personal workspace. It is strongly recommended to use as input the output file created by “Transformation to Equivalent Dimensions” Application.
• Number of clusters – The User selects the number of cluster within which the entire data sample will be divided. The number must be an integer greater than 1.
• Clustering algorithm – The User may select between 2 clustering algorithms, namely “Partitioning” clustering (K-means algorithm) or “Hierarchical” algorithm (Ward’s method). Fuzzy clustering is not supported yet in IS-EPOS platform.
• Clustering method – The User may now select the method for the clustering tree generation, among “average”, “centroid”, “complete”, “median”, “single”, “ward” and “weighted” (this option is only available when “Hierachical” Clustering algorithm is selected).
• Number of leaf nodes – The User finally selects the number of leaves to be demonstrated the final Ward diagram. The number must be an integer greater than 1 and smaller than the total number of the data points (this option only is only available when “Hierachical” Clustering algorithm is selected).

After defining the aforementioned parameters, the User shall click on the  ‘RUN’ button (green tab in Figure 4) and the calculations are performed. The Status changes from 'CREATED' through  'SENT_TO_SERVER', 'RUNNING' and finally 'FINISHED' and the output is created and plotted in the main window. The Analysis Results table appear on the screen and comprise the following outputs:

A) summary report (Figure 2, red frame) with the number of events included in each cluster created together with the center of mass of each cluster in each one of the dimensions in the selected parameter phase space (in Equivalent Dimensions, therefore ranging between 0 to 1). If the phase space has 2 or 3 Dimensions (2 or 3 parameters  have been selected for cluster analysis), then an additional 2D or 3D plot in the corresponding phase space is also created (Figure 2). The summary report is also available and can be downloaded as mat file (cluster_analysis_out_cluster.mat)

Figure 2. Output summary and 3D plot created by “Cluster Analysis” Application

B) A matlab structure (mat) with the transformed parameters of each event together with an index of the cluster to which each event belongs.

C) If Clustering Algorithm “Hierarchical” is selected, a Ward diagram is produced as well showing the linkage distance among the events and the number of clusters created (Figure 3).

Figure 3. Output Dendrogram created by “Cluster Analysis” Application, after “Hierarchical” clustering algorithm was selected.