Mean Shift Clustering
Usage Instruction
Mean shift clustering is a clustering algorithm based on iterative solution of cluster centers, which differs from K-means clustering in that it does not require prior determination of the number of clusters K. The principle of mean shift clustering is to start from the initial cluster center and keep searching for new cluster centers in the direction of increasing density of point groups in the bandwidth until the termination condition is satisfied. The termination condition is that the distance between the centers of the two clusters is almost constant or the maximum number of iterations (default number of iterations: 300) is reached during the iteration.
Parameter Description
| Parameter Name | Default Value | Parameter Definition | Parameter Type |
|---|---|---|---|
| Source Datasets | Set the vector dataset to be analyzed, supporting the point dataset. | DatasetVector | |
| Target data source | The data source where the specified dataset of stored results is located。 | Datasource | |
| Resulting dataset name | Name of the specified result dataset。 | String | |
| Bandwidth | -1 | The bandwidth value is the search radius when a cluster centroid is the core, the unit is consistent with the coordinate system, the value is greater than 0, the default value is -1, the system will automatically calculate a default bandwidth value. The clustering result is affected by the bandwidth. If the bandwidth is set too small, the convergence will be slow and the number of clusters will be too many; if the bandwidth is set too large, some clusters may be lost. | double |
| Custom initialized clustering center seed points | Customize the initial clustering center. Not set by default. The filled form is: [{"x":0.0, "y":0.0},{"x":9.0, "y":9.0}] . | Point2Ds |
Output Result
- A "Cluster_ID" field is added to the source dataset to indicate the resultant clustering categories;
- The resultant vector dataset indicates the cluster centroids of the final clusters.