Prac cal 5a

Step 1: Launch Weka

1. Open Weka - Download and install Weka from Weka’s official website if you haven’t already.

2. Start Weka - Once installed, open the Weka GUI Chooser.

Step 2: Load the Dataset

1. Choose the Explorer Op on - In the Weka GUI Chooser, select Explorer.

2. Open Dataset - Click on Open file… to load a dataset. Weka supports. arff, .csv, and other
formats.

o Sample Datasets - If you don’t have a dataset, Weka includes sample datasets. You
can find these under the data directory within your Weka installa on. Popular
examples include iris.arff or weather.arff.

Step 3: Preprocess the Data

1. Examine the Dataset - In the Preprocess tab, you can view a ributes, instances, and
summary sta s cs of the dataset.

2. Filter and Edit (if needed) - You can use filters to remove or transform a ributes, but for
building a basic decision tree, this step is op onal.

Step 4: Select the Decision Tree Algorithm

1. Go to the Classify Tab - Switch to the Classify tab in the Explorer.

2. Select Classifier - Click Choose to select a classifier.

o Decision Tree - Under trees, choose J48 (the Weka implementa on of the C4.5
algorithm for decision trees). This is the default decision tree algorithm in Weka.

Step 5: Configure the Decision Tree Parameters (Op onal)

1. Set Parameters - A er selec ng J48, you’ll see its default se ngs in the classifier box. You
can change se ngs by clicking on the text box next to J48.

o Confidence Factor - This controls pruning, which reduces overfi ng by simplifying

the tree. The default is 0.25.

o MinNumObj - Sets the minimum number of instances per leaf. Higher values result
in smaller trees.

o Unpruned - If selected, the tree will not be pruned, which may increase complexity
and overfi ng.

2. Confirm Se ngs - Once you’ve configured the parameters (or le them at default), click OK.
Step 6: Choose the Class A ribute

1. Select Class A ribute - At the bo om right, make sure the Class a ribute (the target
variable) is correctly set. This is typically the last a ribute in the dataset, but you can change
it if necessary.

Step 7: Set Up Evalua on Method

1. Cross-Valida on - The default evalua on is 10-fold cross-valida on, which is generally

suitable for most scenarios. This splits the data into 10 parts, trains on 9, and tests on 1,
repea ng for all combina ons.

o Alterna vely, you can choose Percentage split (e.g., 66% training, 34% tes ng) or
Use training set to evaluate the model.

Step 8: Build the Model

1. Click Start - Once everything is set up, click Start to build the model. Weka will train the
Decision Tree on the dataset and display the results.

Step 9: Analyze the Results

1. View Model Output - A er training, Weka will display results in the Classifier output sec on.

o Summary - You’ll see evalua on metrics such as accuracy, precision, recall, F1 score,
and confusion matrix.

o Decision Tree Visualiza on - To see the tree, click on Visualize tree at the bo om.
This gives a graphical representa on of the decision tree.

Step 10: Save the Model (Op onal)

1. Save Model - To save your model, right-click on the model in the Result list on the le pane,
then select Save model.

2. Save Predic ons (Op onal) - You can also save the predic ons by right-clicking on the model
and selec ng Save result buffer.

Step 11: Use the Model for Predic on (Op onal)

1. Load New Data - You can use the Supplied test set op on in the Classify tab to load a new
dataset for predic on.

2. Predict - Once the new data is loaded, click Start to classify instances in the new dataset
using the trained decision tree.
 Prac cal 5b.
Step 1: Launch Weka

1. Open Weka - Open the Weka applica on on your computer.

2. Choose Explorer - In the Weka GUI Chooser, select Explorer.

Step 2: Load the Dataset

1. Open File - In the Explorer, go to the Preprocess tab and click on Open file… to load your
dataset.

2. Supported Formats - Weka supports .arff, .csv, and other file formats. You can select any
dataset you want to classify using Naïve Bayes.

o Sample Datasets - Weka includes sample datasets. You can use iris.arff or
weather.arff (available in Weka’s data folder) if you don’t have a dataset.

Step 3: Preprocess the Data (Op onal)

1. Check A ributes - In the Preprocess tab, you can view all the a ributes and their data types.

2. Filter Data - If you need to clean or transform data, you can use Filters here. This is op onal
for a basic Naïve Bayes model.

Step 4: Choose the Naïve Bayes Algorithm

1. Go to the Classify Tab - Click on the Classify tab to move to the classifier sec on.

2. Select Naïve Bayes - Click Choose, go to bayes, and select NaiveBayes.

o NaiveBayesMul nomial - You might see op ons like NaiveBayesMul nomial, which
is suited for text classifica on and data with nominal a ributes. For a general
numeric and nominal dataset, choose the standard NaiveBayes op on.

Step 5: Set Up Evalua on Method

1. Choose Evalua on Technique - By default, Weka uses 10-fold Cross-Valida on, which is
o en effec ve for classifica on tasks.

o Cross-Valida on - The dataset is divided into 10 parts. The model is trained on 9

parts and tested on 1 part, repea ng for all parts.

o Percentage Split - Alterna vely, you can select Percentage split (e.g., 66% training,
34% tes ng).

o Use Training Set - You can also evaluate the model on the en re training dataset,
although this approach doesn’t give a realis c es mate of performance on unseen
data.

Step 6: Build and Run the Model

1. Set Class A ribute - Ensure the target Class a ribute (the one you want to predict) is
correctly specified at the bo om-right of the window.

2. Click Start - With everything set, click Start to train and evaluate the Naïve Bayes model on
your data.

Step 7: Analyze the Results

1. View Model Output - A er running the model, Weka will display results in the Classifier
output sec on.

o Summary - Check key metrics such as accuracy, precision, recall, F1 score, and the
confusion matrix to understand the performance of the classifier.

o Detailed Accuracy by Class - This provides precision, recall, F-measure, and ROC area
for each class.

o Confusion Matrix - This shows the number of correct and incorrect predic ons for
each class.

Step 8: Visualize Results (Op onal)

1. Visualize Errors - To see the misclassified instances, click Visualize classifier errors at the
bo om.

2. Plot Results - You can also generate 2D visualiza ons of a ribute distribu ons or
classifica on boundaries by selec ng Visualize.

Step 9: Save the Model (Op onal)

1. Save Model - Right-click on your model in the Result list (le side of the Classify tab) and
choose Save model to save the Naïve Bayes model for future use.

Step 10: Make Predic ons on New Data (Op onal)

1. Load New Data - If you want to classify a new dataset with your saved Naïve Bayes model,
load the new data in the Preprocess tab.

2. Classify New Instances - Go to the Classify tab, load your saved model, and classify the new
data.

Using these steps, you can apply Naïve Bayes on any dataset in Weka to perform
classifica on.