Attention:
- Do not edit this file in text editors like Word. Use a plain text editor only. In case of doubt, you can use Spyder as a text editor.
- Do not change the structure of this file. Just fill in your answers in the places provided (After the R#: tag).
- You can add lines in the spaces for your answers but your answers should be brief and straight to the point.
QUESTIONS:
Q1: Considering the data provided, explain the need to standardize the attribute values.
R1:
Q2: Explain how you calculated the parameters for standardization and how you used them in the test set.
R2:
Q3: Explain how you calculated the prior probability of an example belonging to a class (the probability before taking into account the attribute values of the example) in your Naïve Bayes classifier implementation. You may include a relevant piece of your code if this helps you explain.
R3:
Q4: Explain how your Naïve Bayes classifier predicts the class to which a test example belongs. You may include a relevant piece of your code if this helps you explain.
R4:
Q5: Explain the effect of the bandwidth parameter on your classifier.
R5:
Q6: Explain what effect the gamma parameter has on the SVM classifier.
R6:
Q7: Explain how you determined the best bandwidth and gamma parameters for your classifier and the SVM classifier. You may include a relevant piece of your code if this helps you explain.
R7:
Q8: Explain how you obtained the best hypothesis for each classifier after optimizing all parameters.
R8:
Q9: Show the best parameters, the estimate of the true error for each hypothesis you obtained (your classifier and the two provided by the library), the ranges in the expected number of errors given by the approximate normal test, the McNemar test values, and discuss what you can conclude from this.
R9:
Q10: (Optional) Show the estimate of the true error of the optimized SVM classifier (if you did the optional part of the work) and discuss whether it was worth doing this optimization. If you did not do the optional part leave this answer blank.
R10: