Show Menu


  1. 打开 Cluster Builder .
    单击​ 可视化 > 预测分析 > 聚类 > 聚类生成器
  2. 选择输入变量。
    • Add metrics to the Input Variables list by selecting from the Metric menu in the toolbar.
    • Add dimension elements to the Input Variables list by dragging them from a Dimension's table.
      Press Ctrl + Alt and drag selected dimension elements to the Input Variables list or to the Element box in the toolbar.
      默认情况下,聚类会在整个数据集上执行。You can see all input variables in the left Preprocessing pane.
  3. Use the Options menu to select the desired number of clusters.
  4. 如果您要在数据集中聚集访客数的子集,则可以定义人群过滤器。
    Start by defining the desired subset using selections in your Workspace or by using the Filter Editor . Once you have the desired subset selected, set the Target Population in the Options menu. 建议您为目标群组提供一个标识名称。
    The Options menu also has settings to control the maximum number of passes and the acceptable threshold for center convergence.
  5. After inputs and options have been configured, click the Go button to run the clustering locally or press Submit to send the task to the Predictive Analytics Server. 收敛完成时,提交到服务器的任务会将结果维度保存到数据集。
    当在本地运行时,您将看到聚类生成器会随其根据输入定义智能中心,在四个 Canopy 聚类阶段之间移动。
  6. 自定义聚类。
    量度输入为每个聚类提供一个 T 检定,而维度元素输入为每个聚类提供三个分配测试(卡方、熵 U 统计及 Cramer's V 统计)。
    If you add or remove inputs during convergence, the process will pause until you press Go again.
  7. 聚类维度收敛后,您可以将量度添加到表格中,并像往常一样做出选择。您还可以右键单击元素名称(聚类 1、聚类 2 等)来打开上下文菜单,将这些元素重命名为更有意义的名称。
  8. If you wish to use this cluster dimension in other visualizations, you can Save it locally or Submit it to the server.
When selected, Reset will completely release all the input variables and give you a blank cluster builder visualization to define new clusters.