Demonstrate that you evaluated the data set and applied aduquate preprocessing to the data

Learning Goal: I’m working on a artificial intelligence question and need an explanation and answer to help me learn.

Problem 1 – Decision Trees (20 points)

• • Pull the data from https://archive.ics.uci.edu/ml/datasets/Credit+Approval
  • Create a decision tree to determine if credit should be extended based on a test case.
  • Grading criteria:
    • Demonstrate that you evaluated the data set and applied aduquate preprocessing to the data
    • Make sure you comment you code and the cleaning process so we can follow your logic in grading
    • Provide a confusion matrix for your results. Text based is fine.
    • Provide a visualization with explanation that demonstrates logical evaluation of the model
  • Actual accuracy can depend on how you split the training and test data and other random variations
    • If you get below 70% accuracy, there may be a problem with your model
  • Spoiler Alert: If you don’t start with some exploration to determine how to approach data cleaning, this will be more difficult than it should be.
  • Put the explanation of your model here:

PROBLEM 2 – K-MEANS (10 POINTS)

• Use the Ecoli dataset at https://archive.ics.uci.edu/ml/datasets/Ecoli
• Ignore the label and create clusters using k values between 4 and 6.
• Pick the best k value and explain why you picked it
• Show any calculations you used to pick the best cluster
• Create two visualization
  • One colors the nodes with the cluster membership
  • The other colors the nodes based on the actual label
• Grading criteria: Adequately describe how to pick the best cluster and successful create the required visualization
• Provide an explanation of your model:

PROBLEM 3 – SUPPORT VECTOR MACHINES (10 POINTS)

• Use the Iris trainging set
• Explore the data to find the best two features to use
  • We are mostly doing this so we can visualize the results
• Split the data set into 80% training and 20% testing
• Create a SVM to model the data
• Create a visualization that shows the line and the margins
• Create anonther visualization that shows the decision surface
  • Do not include the test data points
  • Randomly select 10 test points and add them to the visualization. Color them based on their label
  • Are the random test points consistently on the correct side of the line?
• Predict the label for ALL of the test data
  • Show a confusion matrix
  • Calculate the F1 measure
• Grading criteria:
  • SVM graphically appears to correctly to use a reasonable line
  • F1 measure is consistent with what we showed in class
  • Explanation of your model: