• Welcome to AppraisersForum.com, the premier online  community for the discussion of real estate appraisal. Register a free account to be able to post and unlock additional forums and features.

How does AI class a house?

OSU Beavers

Elite Member
Joined
Jan 10, 2007
Professional Status
Licensed Appraiser
State
Oregon
How does AI class a house? Assuming that there is enough information from your photos and that the AI correctly figures the ceiling height, number of fixture, types of materials, etc., does AI use a check list, a points system or something else?
 
How does AI class a house? Assuming that there is enough information from your photos and that the AI correctly figures the ceiling height, number of fixture, types of materials, etc., does AI use a check list, a points system or something else?

If you use RCA:

1. You create a statistical model using primarily MARS (or R/earth). If you are one of a handful of advanced appraisers, you might feed the R/earth output into glmnet or mgcv to get an improved model.
2. Then you let your software predict the prices of all comps (and the subject) based on measured variables.
3. Then subtract those estimates from the actual sales prices to ge residuals.
4. Then rank all comparables by residual. The residual is the measure of the value contribution of all subject features such as condition, quality, **, aesthetics and so on - or in other words - the "sex appeal" of the house. That ranking of the residuals is the Market telling which homes, or which sets of subjective featurs, have the most and least value. It is the MARKET setting the rules.
5. You study the ranking and conclude that the MARKET says this kind of home is more vaulable in appeal than another.
6. Fit the subject in that ranking between Fixers on the low end and Beauties on the high end.

So, in conclusion, it takes an experienced appraiser plus AI to do the job. The AI uses the MARKET to set the rules, and human intelligence is very good at beauty contest-type rankings. It is absolute objectivity combined with human intelligence and judgement. Constraints manage things in the background.
 
I am trying to find a non-biased way to make a call for those borderline homes.

I figure that a points system would be the most fair.
 
Percentage worse or better. Aka ranking. That is the way to go. Forget absolute classes like C1, C2, C3, C4, C5, C6. They are subject to wild interpretation in many cases. When you say House A is better than 60% of the other houses in the Mkt Area, - I pretty much know what you mean and if I know that market area, will most likely agree or disagree with you on the spot. Tell me that is better than 99%, 90%, 80%, 70%, ... yes I have. very good idea what that means. Of course, you will need to give me an exact definition of said Market Area.
 
Percentage worse or better. Aka ranking. That is the way to go. Forget absolute classes like C1, C2, C3, C4, C5, C6. They are subject to wild interpretation in many cases. When you say House A is better than 60% of the other houses in the Mkt Area, - I pretty much know what you mean and if I know that market area, will most likely agree or disagree with you on the spot. Tell me that is better than 99%, 90%, 80%, 70%, ... yes I have. very good idea what that means. Of course, you will need to give me an exact definition of said Market Area.
That is how you get steps like "three months ago you called this sale a C3, but in this report you classed it as a C4" Bad appraiser. Most of the time the explanation is that I had more accurate information for the second report, but still it would be nice to AI check some of my calls.
 
That is how you get steps like "three months ago you called this sale a C3, but in this report you classed it as a C4" Bad appraiser. Most of the time the explanation is that I had more accurate information for the second report, but still it would be nice to AI check some of my calls.
I see the problem as that in many market areas, most of the homes are C4, and there is no differentiation. Yet if, instead, you used a scale of 0.00-10.00 based on percentage for worse condition, you would see a significant difference in market value for C4 homes. That is from the objective ranking of MARS residuals. Of course, no doubt in the lower and upper 20%, you can see far more dramatic price increases. But you can't expect regression to give you a good adjustment based on C1-C6. On top of that, there are many homes that are in a grey area between C definitions. Same for Q.
 
Fannie failed with the UADs.
The C ratings were suppose to be absolute that even a not too smart appraiser should figure out the rating.
Instead of making the form long and tedious, Fannie should have gone back and resolve the UAD ratings so even a trainee could figure it out.
 
Extracting Effective Age would serve to rank houses by the inverse of their estimated total lives. With 50 year TEL, that's 2% increments.
 
Sorry I got my C & Q mixed up. I am looking for a non-bias way to judge those borderline Q4 to Q3 houses.
 
Last edited:
Find a Real Estate Appraiser - Enter Zip Code

Copyright © 2000-, AppraisersForum.com, All Rights Reserved
AppraisersForum.com is proudly hosted by the folks at
AppraiserSites.com
Back
Top