Scribd Logo
Upload

Welcome to Scribd!

  • 91 views

    Practical 5: Introduction To Weka For Classfication

    Uploaded by

    Phạm Hoàng Kim
    This document provides instructions for using the Weka data mining software to classify weather and bank customer data using decision trees. It describes loading datasets into Weka, exploring the data attributes, using ID3 and J48 classifiers with different test options, and interpreting the results including decision tree accuracy and structure. Key steps include comparing classifier accuracy using different test options, explaining why training and testing on the same data is invalid, identifying useless attributes, and making predictions from the generated decision trees.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Practical 5: Introduction To Weka For Classfication

    Uploaded by

    Phạm Hoàng Kim
    91 views4 pages
    This document provides instructions for using the Weka data mining software to classify weather and bank customer data using decision trees. It describes loading datasets into Weka, exploring the data attributes, using ID3 and J48 classifiers with different test options, and interpreting the results including decision tree accuracy and structure. Key steps include comparing classifier accuracy using different test options, explaining why training and testing on the same data is invalid, identifying useless attributes, and making predictions from the generated decision trees.

    Original Title

    CSCU9YE_Lab5DecisionTrees_a16

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document provides instructions for using the Weka data mining software to classify weather and bank customer data using decision trees. It describes loading datasets into Weka, exploring the data attributes, using ID3 and J48 classifiers with different test options, and interpreting the results including decision tree accuracy and structure. Key steps include comparing classifier accuracy using different test options, explaining why training and testing on the same data is invalid, identifying useless attributes, and making predictions from the generated decision trees.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    91 views4 pages

    Practical 5: Introduction To Weka For Classfication

    Uploaded by

    Phạm Hoàng Kim
    This document provides instructions for using the Weka data mining software to classify weather and bank customer data using decision trees. It describes loading datasets into Weka, exploring the data attributes, using ID3 and J48 classifiers with different test options, and interpreting the results including decision tree accuracy and structure. Key steps include comparing classifier accuracy using different test options, explaining why training and testing on the same data is invalid, identifying useless attributes, and making predictions from the generated decision trees.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 4

    University of Stirling

    Computing Science and Mathematics


    CSCU9YE, Artificial Intelligence

    Practical 5: Introduction to Weka for Classfication


    Nadarajen Veerapen and Gabriela Ochoa

    Introduction to Weka

    1. Download weather.nominal.arff. A small dataset with attributes describing weather


    conditions, and a decision of whether it is desirable to play outdoor or not.
    2. Open Weka and choose Explorer.

    3. Load weather.nomimal.arff (Open file… button)

    1
    4. Have a look at the different attributes. In Current relation, we can see that there are 14
    instances and 5 attributes in the dataset. Click on each attribute to see its properties in
    Selected attribute and a graph of the distribution of the values of the attribute. The colours
    in the graph each correspond to a class. Pay attention to the type of the attributes. In this
    dataset all the attributes are Nominal: the values indicate different distinct categories that
    describe the attribute. An attribute could also be Numeric: the values are numbers that
    measure the attribute. Notice that play has been suggested as the class attribute, that is the
    one that is predicted from the other attributes.
    5. Have a look at the data (Edit… button). We can see all the data in this window. Each of the
    rows corresponds to an instance and the columns are the attributes.

    6. We are now going to build a decision tree. On the Classify tab, the default classifier is ZeroR
    so click on Choose to select the Id3 classifier from the trees.

    2
    ID3 is one of the simplest decision tree classifiers. Clicking on the classifier name text box, in
    this case Id3, will bring up a window providing a very short description of the classifier. Click
    on More for a bit more details and on Capabilities to know the kinds of attributes and classes
    the classifier can handle.

    This information tells us that the ID3 algorithm can only handle nominal attributes and
    cannot deal with missing values. We can therefore apply it to our data. Note that classifiers
    that are not compatible with the data are greyed out and cannot be selected.
    7. The Test options allow us to choose how to train the classifier.
    a. Use training set will use all the data for both the training and the test sets.
    b. Supplied test set allows you to provide a separate test set.
    c. Cross-validation will perform cross-validation according to the number of folds
    provided. This means that the data will be split into k subsets of equal size. For each
    value of i in {1, 2, … , k} the classifier will be tested on the ith subset after being
    trained on all the other data. The k results are then averaged to describe the
    performance of the classifier.
    d. Percentage split will train the classifier on the indicated percentage of the data and
    test it on the rest.
    8. The dropdown menu allows us to choose the class attribute. Here play has already been
    appropriately suggested. Clicking Start will execute the training and evaluation process. For
    the moment, let us just try Use training set and click Start.
    9. The results are displayed in the Classifier output panel. outlook
    a. The decision tree is given in text form:
    sunny rainy
    outlook = sunny
    | humidity = high: no overcast
    | humidity = normal: yes humidity yes windy
    outlook = overcast: yes high normal TRUE FALSE
    outlook = rainy
    | windy = TRUE: no
    no yes yes
    | windy = FALSE: yes no
    and corresponds to the representation on the right.
    Unfortunately, in Weka, we cannot see a visualisation of a tree produced by ID3.
    However, this is possible for the J48 classifier, which is an implementation of the
    C4.5 algorithm. To visualise a tree, right-click on the corresponding result in the
    Result list and choose Visualize tree.
    b. A summary of the evaluation gives information such as the percentage of correctly
    and incorrectly classified instances.

    3
    c. Information about the accuracy is given. TP and FP refer to True Positives and False
    Positives respectively.

    d. The confusion matrix contains information about the prediction in terms of true and
    false positives and true and false negatives.
    Prediction
    Actual # true positives # false negatives
    value # false positives # true negatives

    You can find lots of information about classifier accuracy and confusion matrices
    online, for example http://www.dataschool.io/simple-guide-to-confusion-matrix-
    terminology

    Classification

    1. Open the weather.nominal dataset.


    a. Compare the accuracy of the J48 classifier when tested using the Use training
    set option and when using different values of Percentage split.
    CHECKPOINT b. Suggest an explanation for the difference in accuracy. Why is it not good
    practice to use the same instances for training and testing?
    2. Open the bank-data dataset which contains data about a bank’s customers and whether
    or not they bought a Personal Equity Plan (PEP is the class).
    a. Inspect the actual data in the dataset (Edit button on Preprocess tab), paying
    attention to the type of the attributes and to attributes where each instance has
    a different value.
    CHECKPOINT b. Can ID3 be used as a classifier on this dataset? Why?
    c. Which attribute is useless to predict the class? Select it and remove it
    (Preprocess tab).
    d. Use the J48 classifier with a Percentage split of 50%. Examine the results and
    generate the visualisation of the tree. If the visualisation window is too small,
    maximise it, right-click in the window and choose Fit to Screen.
    CHECKPOINT e. Based on the prediction from the decision tree:
    i. Is an unmarried customer without children and no mortgage likely to
    buy a PEP?
    ii. What about a customer with more than one child and an income of less
    than £30000?

    You might also like