Assume ?=.05 unless otherwise indicated . A tax assessor wishes to develop a mod
ID: 3918662 • Letter: A
Question
Assume ?=.05 unless otherwise indicated.
A tax assessor wishes to develop a model to be used for estimating the value of a home for determining property taxes. A local real estate firm provided a data set (columns C6 thru C21 of exam2.mtw) in a particular development area (location, location, location) containing the predictor variables:
Predictor Variables
1) Value: Most recently assessed home value (Y=Value)
2) Acreage: Area of lot in acres
3) stories: number of stories the home has
4) Area: Square footage
5) Exterior: 1=Exterior in excellent or good condition, 0=Exterior in average or poor condition.
6) NatGas: 1= has natural gas heat (the most desirable), 0 = other heating system
7) Rooms: total number of rooms
8) Bedrooms: Number of bedrooms
9) FullBath: number of full bathrooms
10) Halfbath: number of half bathrooms
11) Fireplace: 1=Yes 0 = No
12) Garage: 1=Yes 0=No
13) Area**2: Area of home squared
14) Acreage**2: Acreage of lot squared
15) Stories**2: number of stories the home has squared
16) Rooms**2: total number of rooms squared
a.) Based on these Best Subsets results, use the various selection criteria discussed in class to recommend two or 3 potential models for further evaluation.
b.) Discuss the limitations of this Best Subsets selection algorithm. Why shouldn’t you just recommend a single model and be done with it?
c.) Based on your potential models, and additional analysis, recommend a model. Discuss why you chose this particular model.
Predictor Variables
1) Value: Most recently assessed home value (Y=Value)
2) Acreage: Area of lot in acres
3) stories: number of stories the home has
4) Area: Square footage
5) Exterior: 1=Exterior in excellent or good condition, 0=Exterior in average or poor condition.
6) NatGas: 1= has natural gas heat (the most desirable), 0 = other heating system
7) Rooms: total number of rooms
8) Bedrooms: Number of bedrooms
9) FullBath: number of full bathrooms
10) Halfbath: number of half bathrooms
11) Fireplace: 1=Yes 0 = No
12) Garage: 1=Yes 0=No
13) Area**2: Area of home squared
14) Acreage**2: Acreage of lot squared
15) Stories**2: number of stories the home has squared
16) Rooms**2: total number of rooms squared
Explanation / Answer
a) The 3 potential "regression" models could be:
1)Decision trees
2)Random Forests
3)Gradient decent
4) Using techniques like Bucketing,Boosting (as in "Gradient Boosted Decision Trees") etc to combine multiple models for predictions
b)Using multiple models is usually better because it helps in reducing the error in our predictions. Moreover we can utilise the advantages of different models to imprive the accuracy of our predictions.
c) I would use "Gradient Boosted Decision Trees' because it provides a very high accuracy by reducing the error using multiple models...
Related Questions
drjack9650@gmail.com
Navigate
Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.