WEKA SOLUTION

The Experiment was set up using WEKA tool for datamining. The experiments carried out saw the to the development of two predictive models the first one being the Bayesian Network for classifying the dataset while the second one is the optimized Bayesian Network for classifying the same dataset using Genetic algorithm for feature selection i.e. identifying the best set of attributes for Bayesian Network to make correct classification.

The Experiments were carried out and the performance of the learned networks were measured using metrics such as accuracy (%) of correctly classified instances, incorrectly classified instances (%), kappa statistics, Precision, Recall and Receiver Operating Characteristics (ROC) upon a 10-fold cross-validation test mode. Also, the time taken to build the model was recorded in seconds.

Step – by – Step Procedure for Repeating the ‘Bayesian Network Optimized using Genetic Algorithm’ using Weka:

Having install WEKA version 3.8 and also Genetic search algorithm via package manager,

1. Open the WEKA software program

2. Click on the ‘Explorer’ button

3. From the new window, click on the open file button and locate the Iris dataset on your PC

4. Navigate to the ‘Classify’ tab on the panel, click on the choose button

5. Select ‘Meta’, then ‘AttributeSelectedClassifier’

6. Click on the ‘AttributeSelectedClassifier’, from the dialogue box click on the ‘choose’ button in front of ‘classifier’ option, then select ‘bayes’ folder and later ‘BayesNet’

7. Click on the ‘choose’ button in front of the ‘search’ option then select ‘GeneticSearch’

8. Then click on the ‘OKAY’ button

9. The Test option is set to 10fold cross-validation by default

10. The class label is selected by default too, which is always the last attribute, in this case named ‘class’ of nominal value

11. Then click on the ‘START’ button, which kick off the execution of the algorithm and output run information, evaluation of results into the console.

However, to perform the experiment for Bayesian Network without the optimization of genetic search, starting from ‘iv’ above,

v Select ‘bayes’ folder and later ‘BayesNet’

vi Then click on the START button.

You can repeat to same steps for BaysianNet module.

Figure 1:

As it is depicted above, the model of Bayesian Network that was optimized using Genetic Algorithm had a better classification accuracy and also a lower inaccuracy.

Figure 2

As it seen from the Fig 2, Bayesian Networks optimized by Genetic algorithm had better statistical test scores than its counterpart.