Midterm 2....

Question 1

Convert odds of 1:8 to a probability.

Answer 1

1/8 = .125 –> .125/(.125 + 1) = 11.11% probability. Because 1:8 means 1 out of every 9.

Question 2

Odds EQ for Probability(for/against)

Answer 2

odds(for/against)/odds(for/against) + 1

Question 3

What does a logistic regression model predict?

Answer 3

LogOdds! This will have a range of (-infinity, infinity)

Question 4

How do you convert logOdds to odds?

Answer 4

e^logOdds = odds

Question 5

How do you convert logOdds to probability?

Answer 5

e^logOdds/(e^logOdds + 1)

Question 6

What does logOdds equal in terms of ln(x)?

Answer 6

ln(odds) = logOdds or log(odds) = logOdds

Question 7

What is the range of odds (what are they bound by?)

Answer 7

[0, infinity)

Question 8

What is the range of logOdds (what are they bound by?)

Answer 8

(-infinity, infinity)

Question 9

What type of estimation model is logistic regression, and why?

Answer 9

Class probability estimation model. It is using a numeric value to estimate the probability of a categorical variable! Ex. What is the chance Marc goes to class? 0.3

Question 10

What loss function does support vector machine use?

Answer 10

Hinge loss

Question 11

Hinge loss (loss function)

Answer 11

An instance on the wrong side of the line does not incur a penalty. ONLY when it’s on the wrong side and outside of the margin.

Question 12

Zero-one loss

Answer 12

An instance incurs a loss of 0 for a correct decision and 1 for an incorrect decision.

Question 13

Squared error

Answer 13

Specifies a loss equal to the square of the distance from the boundary. A further instance would have a greater error. Usually used for numeric value prediction rather than classification.

Question 14

Loss function

Answer 14

Determines how much penalty should be assigned to an instance based on the model’s predictive value

Question 15

Finish this sentence. Accuracy of training data is sometimes called…

Answer 15

In-sample accuracy (train) vs. out-of sample accuracy (test)

Question 16

When is logistic regression more accurate vs. decision tree and vice versa?

Answer 16

LR is more accurate with a smaller data set, DT on bigger sets

Question 17

What’s the point of regularization?

Answer 17

It gives a penalty to more complicated models because those are more prone to overfitting.

Question 18

In a confusion matrix what are the column headers? Row headers?

Answer 18

Column: Actual y and n
Row: Predicted y and n

Question 19

False positive

Answer 19

Predicted positive, actual negative

Question 20

False negative

Answer 20

Predicted negative, actual positive

Question 21

True negative

Answer 21

Predicted negative, actual negative

Question 22

True positive

Answer 22

Predicted positive, actual positive

Question 23

True positive rate

Answer 23

True positive / all actual positives (both true and false)

Question 24

False positive rate

Answer 24

False positive / all actual negatives (both true and false)

Question 25

Positive predictive value (PPV)

Answer 25

True positive / all predicted positive (both correct and incorrect)

Question 26

What’s the expected value of a game of roulette? Probability of hitting black = 48%. Bet = $100

Answer 26

EV = (0.48)(100) + (1-0.48)(-100) = - 4

Question 27

What are the two uses for expected value?

Answer 27

1. Inform how to use our classifier for individual predictions.
2. Compare classifiers.

Question 28

Class priors

Answer 28

The proportion of positive and negative instances in your data set. Ex. 40 of 100 people would buy a new car next year if they could. p(p) = .4, p(n) = .6

Question 29

Two critical conditions underlying profit calculations:

Answer 29

1. Class priors

2. Costs and benefits

Question 30

Where is the perfect point on an ROC curve (hint: x axis is FPR, y axis is TPR)

Answer 30

Top left. FPR of 0, TPR of 1

Question 31

How are ROC curves created?

Answer 31

TPR and FPR are found at every cutoff point. Ex. Titanic threshold of 0 to 1, values would be found at every .01

Question 32

What is the Area Under the ROC Curve used for (AUC)?

Answer 32

AUC is used when a single number is needed to summarize performance or when nothing

Question 33

What are two alternatives to the ROC curve?

Answer 33

1. Cumulative response curve

2. Lift curve

Question 34

How is a lift curve calculated

Answer 34

Cumulative response curve values. y/x

Question 35

How can you calculate cumulative response curve values from lift?

Answer 35

Lift: x axis * y axis. Contact 20% with a lift of 2.5 means it will be .5 on the cumulative response curve.

Question 36

Euclidian distance

Answer 36

Distance formula but using it with attributes between two people.

Question 37

Manhattan distance

Answer 37

Distance using the two axes rather than the hypotenuse

Question 38

Euclidian vs. Manhattan

Answer 38

Euclidian uses hypotenuse of triangle (where two instances are the edges). Manhattan uses two bases.

Question 39

Nearest neighbors

Answer 39

Judge similarity by calculating distance to nearest neighbors and using those results to make a prediction. Ex. 3 nearest neighbors –> 2 no’s, 1 yes. Instance should be a no!

Question 40

How do we give weight to closest neighbors

Answer 40

With similarity weight: Inverse of distance squared –> contribution = sim. weight/(sum of all sim. weights)

Question 41

How do we get to a probability from nearest neighbors

Answer 41

Sum of all “no” contributions = p(no)

Question 42

How do we avoid overfitting the data with nearest neighbors?

Answer 42

By choosing a higher k = # of neighbors!

Question 43

What are the three issues with nearest neighbors?

Answer 43

1. Dimensionality and domain knowledge. Unimportant features might have too much influence over important ones!
2. Fast to train, slow to predict. Prediction requires plotting the entire dataset.
3. Easy to interpret, but no “knowledge” extracted from data.

Question 44

Hierarchical clustering

Answer 44

Consider individual points and distance between them. Ex. Points with a Euclidian distance smaller than x will be clustered.

Question 45

Link function (clustering)

Answer 45

Minimum req. that must be met before an item is clustered.

Question 46

Centroid-based clustering

Answer 46

Decide k (number of centroids) and groups will be made around those. Points are grouped on which centroid they’re closest to. When a point is added the centroid is repositioned.