Space-Time Clustering with Stability Probe while Riding Downhill

Document Type

Article

Publication Date

1-1-2016

Department

Political Science, International Development, and International Affairs

School

Social Science and Global Studies

Abstract

We propose a new data-driven procedure of optimal selection of tuning parameters in dynamic clustering algorithms, using the notion of stability probe. Due to the shape of the stability probe dynamics, we refer to the new clustering stability procedure as Downhill Riding (DR). We study final sample performance of DR in conjunction with DBSCAN and TRUST in application to clustering synthetic times series and yearly temperature records in Central Germany.

Publication Title

SIGKDD Mining and Learning from Time Series (MiLeTS2016)

Share

COinS