This shows you the differences between two versions of the page.
Both sides previous revision Previous revision | |||
en:classification [2019/03/19 00:23] David Zelený [Unsupervised vs supervised classification] |
en:classification [2019/03/19 08:27] David Zelený [Types of classification methods] |
||
---|---|---|---|
Line 13: | Line 13: | ||
===== Types of classification methods ===== | ===== Types of classification methods ===== | ||
- | Simple “classification” of the numerical classification methods is in <imgref classif-classif>. The methods are either **hierarchical** or **non-hierarchical**, depending on whether the resulting groups of samples have a hierarchical relationship (some are more similar than others, which can be displayed by dendrogram) or not. In case of hierarchical methods, two alternatives exist: **divisive** algorithms, which take the whole dataset and cut it into a subsets and these subsets into a smaller subsets, and **agglomerative** algorithms, which start from the level of individual samples and merge them together into larger groups a little bit like a snowball. | + | Simple “classification” of the numerical classification methods is in <imgref classif-classif>. The methods are either **hierarchical** or **non-hierarchical**, depending on whether the resulting groups of samples have a hierarchical relationship (some are more similar than others, which can be displayed by dendrogram) or not. In case of hierarchical methods, two alternatives exist: **divisive** algorithms, which take the whole dataset and cut it into subsets and these subsets into smaller subsets (in top-down direction), and **agglomerative** algorithms, which start from the level of individual samples and merge them together into larger groups (bottom-up direction). |
- | <imgcaption classif-classif|Classification of classification methods. The dendrograms beside the panels indicate whether the clusters (groups) are hierarchically or non-hierarchically related. The dashed arrow in the case of hierarchical methods indicates the direction of clustering – from higher to lower hierarchy in case of divisive algorithms and from lower to higher hierarchies in the case of agglomerative algorithms. >{{:obrazky:classification-of-classifications.jpg?direct|}}</imgcaption> | + | <imgcaption classif-classif|Classification of classification methods. The dendrograms beside the panels indicate whether the clusters (groups) are hierarchically or non-hierarchically related. The dashed arrow in the case of hierarchical methods indicates the direction of clustering – from higher to lower hierarchy in case of divisive algorithms (top-down direction) and from lower to higher hierarchies in the case of agglomerative algorithms (bottom-up direction). >{{:obrazky:classification-of-classifications.jpg?direct|}}</imgcaption> |