Some Example Questions

CSC 470 Spring 2004

In the Prism algorithm, given the following training data, what would the final learned concept description be? (to ease grading, if there are any ties that would be resolved “arbitrarily”, always choose the first of equals) (test data is not shown here, use the whole set of data)

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	M	Jr	No
PA	M	Sr	No
PA	F	Fr	Yes
PA	F	So	No
NJ	M	Fr	Yes
NJ	M	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Jr	No
NJ	F	Jr	No
CT	M	So	No
CT	M	Sr	No
CT	F	Fr	No
CT	F	So	No
CT	F	Sr	No

1. Find the start of a rule If ____ the Yes

§ PA 3 / 3 = .50

§ NJ 4 / 2 = .67

§ CT 0 / 5 = .00

§ M 4 / 4 = .50

§ F 3 / 6 = .33

§ Fr 6 / 1 = .86

§ So 1 / 3 = .25

§ Jr 0 / 3 = .00

§ Sr 0 / 3 = .00

§ Choice Year = Fr

2. Refine rule
• Fr & PA 2 / 0 = 1.00
• Fr & NJ 4 / 0 = 1.00
• Fr & CT 0 / 1 = .00
• Fr & M 3 / 0 = 1.00
• Fr & F 3 / 1 = 0.75
• Among perfect, choose highest coverage – Rule learned: IF Year=Fr & State=NJ Then Yes
3. Set aside instances covered and learn new rules

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	M	Jr	No
PA	M	Sr	No
PA	F	Fr	Yes
PA	F	So	No
NJ	F	Jr	No
NJ	F	Jr	No
CT	M	So	No
CT	M	Sr	No
CT	F	Fr	No
CT	F	So	No
CT	F	Sr	No

4. Find the start of a rule If ____ the Yes

§ PA 3 / 3 = .50

§ NJ 0 / 2 = .00

§ CT 0 / 5 = .00

§ M 2 / 4 = .33

§ F 1 / 6 = .15

§ Fr 2 / 1 = .67

§ So 1 / 3 = .25

§ Jr 0 / 3 = .00

§ Sr 0 / 3 = .00

§ Choice Year = Fr

5. Refine rule
• Fr & PA 2 / 0 = 1.00
• Fr & NJ 0 / 0 = 0.00
• Fr & CT 0 / 1 = .00
• Fr & M 1 / 0 = 1.00
• Fr & F 1 / 1 = 0.50
• Among perfect, choose highest coverage – Rule learned: IF Year=Fr & State=PA Then Yes
6. Set aside instances covered and learn new rules

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	So	Yes
PA	M	Jr	No
PA	M	Sr	No
PA	F	So	No
NJ	F	Jr	No
NJ	F	Jr	No
CT	M	So	No
CT	M	Sr	No
CT	F	Fr	No
CT	F	So	No
CT	F	Sr	No

7. Find the start of a rule If ____ the Yes

§ PA 1 / 3 = .25

§ NJ 0 / 2 = .00

§ CT 0 / 5 = .00

§ M 1 / 4 = .20

§ F 0 / 6 = .00

§ Fr 0 / 1 = .00

§ So 1 / 3 = .25

§ Jr 0 / 3 = .00

§ Sr 0 / 3 = .00

§ Tie, with same Pct and coverage, per instructions, choose first of equals Choice State = PA

8. Refine rule
• PA & M 1 / 2 = 0.33
• PA & F 0 / 1 = 0.00
• PA & Fr 0 / 0 = .00
• PA & So 1 / 1 = 0.50
• PA & Jr 0 / 1 = 0.00
• PA & Sr 0 / 1 = 0.00

§ Choice – State=PA & Year = So

9. Refine rule
• PA & So & M 1 / 0 = 1.00
• PA & So & F 0 / 1 = 0.00
• Rule learned: IF State=PA & Year=So & Gender = M Then Yes
10. Now, bring back all data and learn rules for Fanatic = No

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	M	Jr	No
PA	M	Sr	No
PA	F	Fr	Yes
PA	F	So	No
NJ	M	Fr	Yes
NJ	M	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Jr	No
NJ	F	Jr	No
CT	M	So	No
CT	M	Sr	No
CT	F	Fr	No
CT	F	So	No
CT	F	Sr	No

§ PA 3 / 3 = .50

§ NJ 4 / 2 = .33

§ CT 0 / 5 = 1.00

§ M 4 / 4 = .50

§ F 3 / 6 = .67

§ Fr 6 / 1 = .15

§ So 1 / 3 = .75

§ Jr 0 / 3 = 1.00

§ Sr 0 / 3 = 1.00

• Among perfect, choose highest coverage – Rule learned: IF State=CT Then No
11. Set aside instances covered and learn new rules

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	M	Jr	No
PA	M	Sr	No
PA	F	Fr	Yes
PA	F	So	No
NJ	M	Fr	Yes
NJ	M	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Jr	No
NJ	F	Jr	No

13. Find the start of a rule If ____ the No

§ PA 3 / 3 = .50

§ NJ 4 / 2 = .33

§ CT 0 / 0 = .00

§ M 4 / 2 = .33

§ F 3 / 3 = .50

§ Fr 6 / 0 = .00

§ So 1 / 1 = .50

§ Jr 0 / 3 = 1.00

§ Sr 0 / 1 = 1.00

• Among perfect, choose highest coverage – Rule learned: IF Year=Jr Then No
13. Set aside instances covered and learn new rules

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	M	Sr	No
PA	F	Fr	Yes
PA	F	So	No
NJ	M	Fr	Yes
NJ	M	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Fr	Yes

15. Find the start of a rule If ____ the No

§ PA 3 / 2 = .40

§ NJ 4 / 0 = .00

§ CT 0 / 0 = .00

§ M 4 / 1 = .20

§ F 3 / 1 = .25

§ Fr 6 / 0 = .00

§ So 1 / 1 = .50

§ Jr 0 / 0 = 0.00

§ Sr 0 / 1 = 1.00

• One perfect,– Rule learned: IF Year=Sr Then No
15. Set aside instances covered and learn new rules

Home State	Gender	Year	La Salle basketball fanatic (class to be predicted)
PA	M	Fr	Yes
PA	M	So	Yes
PA	F	Fr	Yes
PA	F	So	No
NJ	M	Fr	Yes
NJ	M	Fr	Yes
NJ	F	Fr	Yes
NJ	F	Fr	Yes

17. Find the start of a rule If ____ the No

§ PA 3 / 1 = .25

§ NJ 4 / 0 = .00

§ CT 0 / 0 = .00

§ M 4 / 0 = .00

§ F 3 / 1 = .25

§ Fr 6 / 0 = .00

§ So 1 / 1 = .50

§ Jr 0 / 0 = 0.00

§ Sr 0 / 0 = 0.00

• Choice – So
17. Refine rule
• So & PA 1 / 1 = 0.50
• So & NJ 0 / 0 = 0.00
• So & CT 0 / 0 = .00
• So & M 1 / 0 = 0.00
• So & F 0 / 1 = 1.00
• One perfect,– Rule learned: IF Year=So & Gender = F Then No
18. All No’s are covered, complete rule set learned:
• IF Year=Fr & State=NJ Then Fanatic =Yes
• IF Year=Fr & State=PA Then Fanatic =Yes
• IF State=PA & Year=So & Gender = M Then Fanatic =Yes
• IF State=CT Then Fanatic =No
• IF Year=Jr Then Fanatic =No
• IF Year=Sr Then Fanatic =No
• IF Year=So & Gender = F Then Fanatic =No