25.10.2016 Views

SAP HANA Predictive Analysis Library (PAL)

sap_hana_predictive_analysis_library_pal_en

sap_hana_predictive_analysis_library_pal_en

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>PAL</strong>_KMEDIANS_ASSIGN_TBL:<br />

3.1.9 K-Medoids<br />

This is a clustering algorithm related to the K-means algorithm. Both k-medoids and k-means algorithms<br />

partition n observations into k clusters in which each observation is assigned to the cluster with the closest<br />

center. In contrast to k-means algorithm, k-medoids clustering does not calculate means, but medoids to be<br />

the new cluster centers.<br />

A medoid is defined as the center of a cluster, whose average dissimilarity to all the objects in the cluster is<br />

minimal. Compared with the K-means algorithm, K-medoids is more robust to noise and outliers.<br />

Given an initial set of K medoids: m1, ..., mk, the algorithm proceeds by alternating between two steps as<br />

below:<br />

●<br />

●<br />

Assignment step: assigns each observation to the cluster with the closest center.<br />

Update step: calculates the new medoid to be the center of the observations for each cluster.<br />

The algorithm repeats until the assignments no longer change.<br />

In <strong>PAL</strong>, the K-medoids algorithm supports multi-threads, data normalization, different distance level<br />

measurements, and cluster quality measurement (Silhouette). It does not support categorical data, but this<br />

can be managed through data transformation. The first K and random K starting methods are supported.<br />

Support for Categorical Attributes<br />

<strong>SAP</strong> <strong>HANA</strong> <strong>Predictive</strong> <strong>Analysis</strong> <strong>Library</strong> (<strong>PAL</strong>)<br />

<strong>PAL</strong> Functions P U B L I C 83

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!