Instructions

Decision Tree Builder

Download Decision Tree Builder Application

In order to run this application, you need to have JVM already installed on your computer. Consult the Java website to download the appropriate version for your computer.
Click the link/button provided and save the DecisionTreeBuilder.jnlp file to the desired location on your local computer. Double click the file and say yes to the security box that pops up.
Each time the application updates and this includes the first time you launch the application you need to wait for the update process before proceeding. You also need to allow the digital jar signature, so that, the application can access your local dataset file.

Loading Data

Before building a decision tree, you need to have a dataset. There are three options available to you:

    • OPTION 1: Loading a dataset file: Click the 'File'-'Open' menu, and choose your local dataset file. In order to use this option you need to have prepared your dataset file and have it saved on your PC in advance. If you want to make your own dataset file, use the format detailed below in the Format section of this page. Two sample datasets are available. By clicking the link below you can download them to your local computer.

      1. Example from DecisionTree.NET

      2. Textbook Exercise 4.2

    • OPTION 2: Importing a dataset from Excel: You can import an Excel 2003 (xls) file by clicking the 'File'-'Import from MS Excel file'.  The same datasets provided in the first option can also be downloaded in Excel 2003 format.

      1. Excel for example from DecisionTree.NET

      2. Excel for Textbook Exercise 4.2


    • OPTION 3: Create a new dataset file: Use 'File'-'New' to create your own data set with Rich Editor self-integrated.

Editing Data

The data that has been loaded from the previous step can be edited by using the table on the left side of the main window. The application will immediately update to reflect these changes. If you want to change the data structure, you can use Rich Editor by clicking 'Control'-'Edit with Rich Editor'.

Tree Construction

On the 'Control' menu, you can decide whether to display calculated gain and which attribute computing measure you want to use. The default measure is: Gini gain display.

    • Left-click the tree node to find out the data it contains in the table, and if the node is a branch node.
    • Right-click the tree node to choose the attribute you are going to use for splitting. Branch nodes like root node and internal nodes are shown in blue, leaf nodes are shown in green. In addition, the conflict node will be shown in black.

Saving Data

You can save the current data set shown in the table to your local file for future use.


Issues

  • If the Gain is less than 0.0001, it will does not display.

File Format

  1. First line: names of the attributes, divided by comma.
  2. Second line: data types of the attributes, use Java format definition, now Support String type, Integer type and Double type. Divided by comma.
  3. Third line: the target attribute name.
  4. From the fourth line: add each data object for a line, devided by comma.