Decision Tree Classifier Agent

The AbleDecisionTreeClassifierAgent provides a classification function. The Decision Tree algorithm is used as the classifier model. To use the Decision Tree agent, the user must specify the data source (and meta data in the form of a *.dfn file). Generating a Decision Tree classification agent will:

  1. Create Import beans to be used for storing data used to construct the decision tree model.
  2. Scan through the source data and create the translate filters required to transform and scale the data for input and output to/from the decision tree bean.
  3. Create the decision tree bean and initialize it with the default values received from the agent (metric, discretization, number attributes, number classes, number records), process training data and construct the decision tree, then switch to test mode. 
  4. Create the data connections for the data to flow through the beans.

At the end of these steps, the classifier agent is configured and the decision tree model is constructed according to the Import corresponding to the input file. Now one of the test or run phases can begin, which means that the decision tree bean is connected to the other import bean and starts its classification function. The user can choose to test the decision tree algorithm via Step/Cycle/Run/Halt buttons on the Agent Editor toolbar. Step button will process one test/run example at a time, while Cycle button will make a pass through all the examples in the test/run file. Run button will cycle through the data until the Stop button is pushed. Whenever the Reset button is pushed, all the corresponding statistics (e.g., accuracy, error, total number of tested examples, number of correctly classified examples) are reset to their default values.