ccao-data / model-condo-avm Goto Github PK
View Code? Open in Web Editor NEWAutomated valuation model for all class 299 and 399 residential condominiums in Cook County
License: GNU Affero General Public License v3.0
Automated valuation model for all class 299 and 399 residential condominiums in Cook County
License: GNU Affero General Public License v3.0
This has already been done manually, but should be incorporated to the export script.
Our iasWorld export needs to be reverted back to building-level exports.
Valuations believes we are likely perpetuating inaccurate common area designations for condo parcels and that common areas should not be designated as condo parcels anyhow. We should completely ignore common area as a category when valuing condos in the pipeline.
For whatever reason, DESCRIPTION
and setup.R
are not loading readr
into the library while running export. Name-spacing it resolves the issue.
The condo model currently handles new construction the same way it handles any other property: by using sales to predict the value of unsold properties. This method doesn't work well for new properties because the sales are inevitably higher than similar non-new properties.
We should add flags and/or a separate valuation methodology for new construction condos. Specifically, we should look for new 299 PINs resulting from divisions and 297 PINs that become 299s. We should also flag large YoY drops in building price, as this can be an additional indicator of misvalued new construction.
Note that for the under construction 297 or 299 PINs may not be the same as the final 299 PINs.
We need to edit the dvc.yaml
file to have each stage have the files invoked in the stage as deps.
Valuations is handcrafting real estate submarkets for condos for sales validation. It might be worth using these geographies rather than township when constructing condo strata.
Similar to ccao-data/model-res-avm#24, we need to setup infra for generating a Quarto report for the condo model, uploading it to S3, and linking to it in the pipeline SNS notification. The work here should be basically the same as ccao-data/model-res-avm#62.
The condo model uses recipes
-based imputation for condo strata. Currently, it uses KNN using Gower's distance and a few of the most salient condo features (year built, distance, etc.). We should revisit this method considering the strata features do most of the work in the condo model and there are many condos in the City.
Nonlivables are difficult to assign a value to pre-disaggregation. If we exclude them from aggregation and then only assign them a value using their relative share of a building's total value, this could avoid shifting values for livable units.
Same as the the residential model, we need to switch to loc_tax_municipality_name
and rerun the ingest script so that it's present in the training data.
We have a number of issues in the backlog to make it easier to deploy and run the residential model:
Once these changes have been deployed to the res model and we're feeling confident in their stability and their usefulness, we should replicate them in the condo model as well.
There may be opportunities for factoring out some shared code into a shared composite action, but that would have the downside of requiring us to manage a third repo containing the action that would need to be updated and versioned in order to make any changes. It may be simpler to just duplicate the logic between these two repos, unless A) the logic is nearly 100% identical or B) we realize that we'll want to use such an action in other repos as well.
model-res-avm
has been undergoing significant upgrades for modelling this year. We need to make sure these upgrades are incorporated into the condo model. This includes:
Need to update numbers, dates, features, pipeline improvements, etc for 2024.
It's a little difficult currently to evaluate the precision with which strata are built and imputed. We should add a section to the pipeline report that breaks down sales prices (min, med, mean, max) by strata that are assigned and imputed.
Mirror the changes from ccao-data/model-res-avm#65 and ccao-data/model-res-avm#66 in this repo, so that the condo model dev dependencies are also isolated and the setup step for each pipeline stage gets factored out into a shared script.
We've changed a lot of column names since we last ran the pipeline. We'll need to update instances where they're hard-coded to make sure the pipeline can run for the sake of investigating the importance of characteristics.
In the 2021 model, the CCAO used a few different methods to value parking. Primarily, it used a flat $10K FMV for most spaces. We should revisit this method and replace it with a variable rate or regression-based model.
default.vw_pin_sale will soon be unfiltered by default and we need to use certain conditions to make sure we don't ingest unwanted sales:
AND NOT sale.sale_filter_is_outlier
AND NOT sale.sale_filter_deed_type
AND NOT sale.sale_filter_less_than_10k
AND NOT sale.sale_filter_same_sale_within_365
This repo has undergone major changes over the last year. We need to update readmes to properly reflect those changes.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.