Bachelor Thesis
This thesis studies three different approaches to cluster time series data using the unsupervised pattern recognition method called hierarchical clustering. The underlying data constitute long-term water temperature measurements of several Swiss water bodies and originates from metering stations which are managed by the Federal Office for the Environment in Switzerland.
The goal is to group these stations according to the resemblance of their hydrologic temperature curve over a period of ten years with a ten-minute sampling rate of detail. Stations that exhibit very similar short-term as well as long-term temperature behaviour and evolution over time should be grouped into the same clusters. These clusterings should provide a better understanding of the data heterogeneity received from the various metering stations in Switzerland and support future decisions regarding the integration of new stations.
The first part of this work addresses the characteristics of time series data and surveys the field of pattern discovery techniques. The procedure of hierarchical clustering is explained in detail as it is the chosen technique applied for the cluster analysis of this thesis. Furthermore, four internal cluster validity indexes used to assess the quality of a cluster composition are elaborated.
The main part addresses the applied distance measuring strategies and assesses the quality of the received clustering results. Defining the level of similarity between two data objects is a fundamental concept in pattern recognition disciplines. This thesis elaborates the two shape-based strategies Pairwise Distance and Dynamic Time Warping and the feature-based strategy Discrete Wavelet Transformation. The cluster analyses are generated with different data aggregation levels and linkage methods. Finally, the various clustering approaches are challenged based on a forecast deviation analysis. This facilitates conclusions about the quality of the various cluster compositions in the form of quantifiable measures.
Keywords: Hydrology · Water Temperatures · Hierarchical Clustering · Time Series