INFO 367
Chapter 6 Problems
3/6/17
6.1
a.) Data should be partitioned into two sets, training, and validation. The training data is the data that is used to create a model and the validation data is used to score
...
INFO 367
Chapter 6 Problems
3/6/17
6.1
a.) Data should be partitioned into two sets, training, and validation. The training data is the data that is used to create a model and the validation data is used to score how successful your model is. If you use the same data to train and validate your model, it will misleadingly say that your model is 100% accurate because it is the same data.
b.) MEDV = -26.36486779 + (-0.296892604 *CRIM) + (5.213218698 * CHAS) + (7.861049277 * RM)
c.) MEDV = -26.36486779 + (-0.296892604 *0.1) + (5.213218698 * 0) + (7.861049277 * 6)
MEDV = 20.77173855
Prediction Error: 0.5162297
d.) INDUS, NOX, and TAX are similar and highly correlated
6.2
a.
Males on average spend less than women
b.
Spending vs FREQ seems to have a little linear relationship, but Spending vs Last Update does not seem to be.
c.
II. SPENDING = 7.266336348 + (12.63558207*US) + (50.02826842* FREQ) + (69570.18212* last_update_days_ago) + (14.66013515 *Web order) + (15.02899451* Gender=male) + (11.51998076 * Address_is_res)
III.
A male who made an online purchase in the US that also makes frequent purchases is likely to spend a large amount of money.
IV. Using backward elimination the first predictor to be dropped would be Address_is_res because it has the lowest statistical significance.
V.
Vi.
Vii.
They have a normal distribution. Most of the residuals predicted higher than the actual value.
6.3
a.
It seems that the best predictor for FARE is DISTANCE.
b.
If the flight is South Western it is nearly half the price of a flight that was not South Western.
6.4
a. Top car specifications for predicting price
- Automatic_airco
- Fuel Type
- Color
b.
The model is accurate at predicting price
[Show More]