Update README and default run_id for 2025
#352
Conversation
dfsnow
changed the title
run_id for 2025
| | PUMA Housing Index | [DePaul Institute for Housing Studies](https://www.housingstudies.org/) | | ||
| | School Ratings | [GreatSchools.org](https://greatschools.org/), aggregated to the district level | |
There was a problem hiding this comment.
Both dropped from the model, so I figured it would be confusing to keep them in this table. They are still present in the training/assessment data however.
| @@ -498,7 +496,7 @@ There a few caveats with this approach and with balance testing in general: | |||
|
|
|||
| In addition to the first-pass modeling done by LightGBM, the CCAO also performs a set of simple adjustments on the initial predicted values from the `assess` stage. These adjustments are internally called "post-modeling," and are responsible for correcting minor deficiencies in the initial predictions. Specifically, post-modeling will: | |||
|
|
|||
| 1. Aggregate values for multi-card properties to the PIN level, then disaggregate them back to the card level. A check is used to ensure that the PIN-level assessed value is not significantly greater than the prior year's value. This is needed because often back buildings (ADUs, secondary buildings) will receive a much higher initial value than they are actually worth (since they are not differentiated as ADUs by the model). | |||
| 1. Handle multi-card properties. For PINs with more than 3 cards, this means summing the estimate for each card to the PIN level, then disaggregating back to the card level. For PINs with 3 or fewer cards, the initial prediction of the _largest_ card (plus the building square footage of the remaining cards) is used as the PIN-level estimate. | |||
There was a problem hiding this comment.
Updates for the new multi-card handling introduced by #328.
| @@ -591,9 +597,9 @@ The property characteristics we track can sometimes be incorrect or outdated. Th | |||
|
|
|||
| These errors can cause under- *or* over-assessment. If you believe your property has been misvalued due to a characteristic error or the property characteristics recorded on our website are incorrect. Please [contact our office](https://www.cookcountyassessor.com/contact) to file a property characteristic appeal. | |||
|
|
|||
| ##### Non-Arms-Length Sales | |||
| ##### Non-Arms-Length/Non-Market Sales | |||
There was a problem hiding this comment.
Updating language per the commercial folks suggestions from awhile back.
| @@ -15,7 +15,7 @@ format: | |||
| fig-align: center | |||
| fontsize: 12pt | |||
| params: | |||
| run_id: "2025-01-13-dazzling-kyra" | |||
| run_id: "2025-02-11-charming-eric" | |||
There was a problem hiding this comment.
Here just setting the default run ID to the final model for the year.
There was a problem hiding this comment.
Thanks for this chore! One quick question: Do we also want to update the top-level Prior Models table in the README to include a link to 2025? Not sure whether that table should include the most recent assessment year or not, I'll defer to whatever you think is best.
|
In the past we have not included the current model/most recent assessment in that table. I think it's fine as-is. |
* Update default run IDs * Add 2025 data links * Add 2025 changelog * Update README for 2025
This PR makes minor updates to the README for tax year 2025.