ICS 171 Homework #16

1. (50) Take your patients X and Y as in previous assignments, and 8 more people from the patient.lisp such that 5 have bacterial_meningitis and 5 have viral_meningitis. By hand, construct a decision tree from these 10 examples (use any criteria you like for selecting tests). You may hand write this part, and turn it in to the TA in class. Do not use the attribute CSFPOLY. Do not use an attribute in a subtree if all examples at the subtree have the same values for it.

2. (50) Convert your decision tree to set of Horn Clause rules. The head of each clause will be (diagnosis ?p bacterial_meningitis) or (diagnosis ?p viral_meningitis) and the body should be the conjunction of condition from the leaf to the root). Evaluate the accuracy of the learned rules on all people as follows:

(load "unify")
(load "fast-traced-prove")
(load "patients")  
(store-clauses *sick-people*) 
(load "learned-rules")  ;You create this file with rules from decision tree 
(store-clauses *learned-rules*)  ;you create this variable in learned-rules
(load "learn")
(test-rules 'diagnosis 'correct-diagnosis)

Test-rules is defined in learn and will try your rules on every person, comparing the diagnosis inferred by your rules to the correct diagnosis.

Back to http://www.ics.uci.edu/~pazzani/171-p.html (text only) or http://www.ics.uci.edu/~pazzani/171.html .

Michael Pazzani
Department of Information and Computer Science,
University of California, Irvine
Irvine, CA 92717-3425