| Resumo: | Clustering has been widely used to find homogeneous groups of data in datasets while looking at some specific metric. Several clustering techniques have been developed, each one presenting advantages and drawbacks to specific applications. This work addresses the development of a clustering technique for the creation of Space Models – STICH (Space Models Identification Through Hierarchical Clustering). Space Models are divisions of the space in which the elementary regions are grouped according to their similarities with respect to a specific indicator (value of an attribute). The identified models, which are formed by sets of clusters, point out particularities of the analysed data, namely the exhibition of clusters with outliers, regions which behaviour is strongly different from the other regions analysed. The results achieved with STICH and with the well known k-means algorithm are compared, allowing the validation of the work developed so far in STICH.
|
|---|