Question

The latest version of Weka is needed for these questions. Use Weka to complete classification task on "iris" data (total: 150 records). You will definitely need to download the data by yourself and likely pre-process the data before loading it to the Weka explorer. Complete and following tasks and also answer the questions.

1. Briefly explain how you pre-process the data so that you can successfully load it into Weka Explorer.

2. What is the minimum length and maximum length of petals within this dataset?

3. Use "REPTree" and the default configuration to create a decision tree. Under "Test options", choose "Use training set". Note that your task is to predict the class of species. Show your tree structure.

4. Click on the bar that shows "REPTree" with parameters. Change the value of "noPruning" (True/False).  Then restart the classification task. Show your new tree structure.

5. What is the difference between the results of 3 and 4? Which results gives you better in prediction? Justify your answer.

Answer 1

1) pre-process the data:

We convert the CSV file to Weka ARFF format by providing necessary headers. We, then load the data using the Pre-process tab in Weka Explorer.

@RELATION iris

@ATTRIBUTE sepallength   REAL
@ATTRIBUTE sepalwidth    REAL
@ATTRIBUTE petallength    REAL
@ATTRIBUTE petalwidth   REAL
@ATTRIBUTE class    {Iris-setosa,Iris-versicolor,Iris-virginica}

@DATA

<CSV data from IRIS file goes here>

2) minimum length and maximum length of petals within this dataset

3) "REPTree" and the default configuration to create a decision tree. Under "Test options", choose "Use training set".

4) Click on the bar that shows "REPTree" with parameters. Change the value of "noPruning" (True/False). Then restart the classification task. Show your new tree structure.

5) What is the difference between the results of 3 and 4? Which results gives you better in prediction? Justify your answer.

After we enable the Tree-Pruning option, overfitting is reduced and the accuracy of the resulting tree improves.

Accuracy with default options (NO PRUNING)

Accuracy after enabling TREE PRUNING