The neighborgram extension provides the following nodes:
- Universe Marker
Prepares the input data to be used in mining nodes for parallel universes, such all the Neighborgram nodes. The dialog allows to label certain input columns as members (attributes) for certain universes, i.e. descriptor space. For a general concept of Parallel Universes see here.
This node is a prerequisite if the neihborgram nodes shall be applied to non-numeric input (like for instance distance vector columns or bit vector, e.g. chemical fingerprint descriptors).
- Universe Marker (Apply)
- Apply the universe definition from a universe marker node to a new data set.
- NG Construct & View
- Computes neighborgrams on the input data (possibly prepared with the Universe Marker node) and displays them. It does not offer for an interactive clustering construction. The neighborgram table (available in the view) will contain as many columns as there are marked universes in the input table.
- NG Construct (beta)
- Constructs neighborgrams and outputs a model suitable for the NG Visual Clusterer (beta).
- NG Visual Cluster (beta)
- Consumes the model constructed in the NG Construct node. It allows for an interactive model construction, i.e. the user can select cluster candidates in the neighborgram view, which will constitute the prediction model.
- NG Learner (beta)
- Automatically generates a prediction model (a classifier) based on the input data.
- NG Predictor (beta)
- Applies a prediction model (either learned by the NG Learner or the NG Visual Clusterer node) to a test set.