Getting familiar with Weka

[damirsaleschoice]

• Open weka UI
• Go to Explorer
• Go to Open File...
• Locate the folder with weka examples, (something like /home/saleschoice/weka-3-8-5/data)
• Open any of them (for example weather-nominal)
• Go to Classify tab
• Press choose button
• Select any of them (for example Rules -> Decision table)
• Close the selection dropdown
• Press Start

Experiment, experiment, play around as much as you can !

BONUS (advanced, to test if you learned enough):

• Select the Neural Nets (that is, Multi Layer Perceptron) , let the classifier display the neural net topography with a popup, start training with the topography setting the maximum number of epochs (any high number such as 9999999).

Question: Why do you think some files can work with some algorithms such as Simple Linear Regression while others cannot ?

[sherlockliang888]

To answer the question, I think it's because the goals are different. It depends on whether the data is used for classification or prediction. For some datasets, we are predicting the continuous dependent variables, such as probability, price.., and that's why SLR works. However, when it comes to predicting/classifying discrete outcomes, we should use other classification algorithms, and SLR simply just won't work.

PS. This is so cool! NN is something we are learning right now in school and it is so cool to choose and visualize the layers! There are so many algorithms I have never seen before, it is so exciting to see Weka being such a powerful tool.

[sherlockliang888]

Done