This preview shows pages 1–3. Sign up to view the full content.
More Handling Missing Data
Data Mining
Prof. Dawn Woodard
School of ORIE
Cornell University
1
Outline
1
Announcements
2
Practice Question
3
Handling Missing Data
2
Announcements
Questions?
4
Error Rates
Say we have a classiFer that, if
X
1
=
0, ﬂips a biased coin;
with probability
p
0
the coin is heads. If the coin is heads we
predict
Y
=
1; otherwise, we predict
Y
=
0. If
X
1
=
1, we
ﬂip a different biased coin; with probability
p
1
the coin is
heads. If this coin is heads we predict
Y
=
1, and
otherwise we predict
Y
=
0.
Say that
Pr
(
X
1
=
1

Y
=
1
)=
0
.
3 and
Pr
(
X
1
=
1

Y
=
0
0
.
6. What is the false positive rate of
our classiFer?
What is the false negative rate?
6
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentNaive Bayes
Say for a different problem our training data looks like:
X
1
X
2
Y
a1
1
a0
0
b0
1
1
1
b1
0
?1
1
?0
0
0
0
1
8
Naive Bayes
What are the (frequentist) estimates of
Pr
(
Y
)
,
Pr
(
X
1

Y
)
, and
Pr
(
X
2

Y
)
for naive Bayes?
X
1
X
2
Y
1
0
1
1
1
0
1
0
0
0
1
9
Naive Bayes
Estimate the probability that
Y
=
0, given that
X
1
=
b
and
X
2
=
0.
This is the end of the preview. Sign up
to
access the rest of the document.
 '08
 DAWN

Click to edit the document details