Skip to main content
CARLA HQ
Real Estate & Property Valuation Finance & Banking use casesNACE L68RegressionTarget: sale_price_usdOpenML #42092 ↗

House Sales

The King County House Sales dataset provides transaction records for 21,613 residential properties sold between May 2014 and May 2015 in Greater Seattle. It serves as a gold-standard benchmark for regression models predicting property market values from structural, geospatial, and neighborhood features.

✨ Try in CarlaInstall Chrome Extension ↗100% In-Browser WebGPU • Zero Cloud Upload
21,613Records (Rows)
20Predictive Features
20 / 1Numeric / Categorical
0.0%Missing Value Ratio
111,524RMSE
62.68sWebGPU Mean Split
BUSINESS CONTEXT & OBJECTIVEReal Estate & PropTech (L68)

Business Objective: House Sales

01Business Context

Automated Valuation Models (AVMs) are central to modern mortgage underwriting, real estate brokerage, iBuying platforms, and property tax assessments. In high-velocity urban markets like King County / Seattle, manual property appraisal is cost-prohibitive and slow, creating substantial demand for automated, highly accurate continuous price estimation.

02Analytical Objective

Predict the final residential property sale price (in USD) using 19 structural, temporal, geospatial, and neighborhood characteristics.

03Economic & Decision Impact

Underpricing leads to lost commissions and seller margin erosion, whereas overpricing extends days on market (DOM) and creates mortgage default and liquidity exposure for institutional portfolio buyers. High-accuracy AVMs reduce appraisal turnaround from weeks to seconds and minimize variance in automated loan-to-value (LTV) calculations.

ML Benchmark Narrative

Tabular gradient boosting architectures like LightGBM and XGBoost achieve strong baseline performance (R² ~ 0.88–0.90) by capturing non-linear spatial interactions between latitude, longitude, and King County building grade. In-context foundation models such as TabICL / Carla deliver competitive zero-shot property valuations directly within spreadsheet environments without hyperparameter tuning, significantly lowering the barrier for real estate analysts and underwriters.

Source Origin:King County Department of Assessments / Kaggle / OpenML
Creator:King County Department of Assessments (2014)
License:CC0: Public Domain
FEATURE SPECIFICATION & DATA DICTIONARY

Target & Feature Column Definitions

Exhaustive business definitions, measurement units, roles, and target variables across all 21 columns.

ColVariable / Feature NameRoleTypeUnit / ScaleDescription & Business Meaning
Asale_price_usdSale Price (USD)TARGETNUMUSD
Final realized sale transaction price of the home in US Dollars.
Bproperty_idProperty IDFeatureNUM
Unique administrative parcel identifier assigned to each residential property in King County.
Csale_dateSale DateFeatureCATISO 8601 string
Timestamp indicating when the property transaction occurred (format: YYYYMMDDT000000).
Dbedroom_countBedroomsFeatureNUMcount
Total count of dedicated bedrooms in the house.
Ebathroom_countBathroomsFeatureNUMcount
Total number of bathrooms, where 0.5 represents a toilet and sink (half bath) and 0.75 represents a shower without a tub.
Fliving_area_sqftLiving Area (Sq Ft)FeatureNUMsq ft
Total interior living space measurement of the home.
Glot_area_sqftLot Size (Sq Ft)FeatureNUMsq ft
Total surface area of the land parcel associated with the property.
Hfloor_countNumber of FloorsFeatureNUMfloors
Total levels or stories in the residential building.
Iis_waterfrontWaterfront ViewFeatureNUMbinary indicator
Binary indicator representing whether the property borders a waterfront (1 = Yes, 0 = No).
Jview_ratingView Rating (0-4)FeatureNUMindex (0-4)
Qualitative index from 0 to 4 grading how scenic the property's surrounding view is.
Kcondition_ratingCondition Rating (1-5)FeatureNUMindex (1-5)
Physical state assessment of the property on an ordinal scale from 1 (poor/uninhabitable) to 5 (excellent).
Lconstruction_gradeKing County Construction Grade (1-13)FeatureNUMgrade level (1-13)
Official King County grading scale reflecting construction materials, craftsmanship, and architectural design quality (1-3 poor, 7 average, 11-13 luxury/mansion).
Mabove_ground_sqftAbove Ground Area (Sq Ft)FeatureNUMsq ft
Square footage of interior living housing space situated entirely above ground level.
Nbasement_sqftBasement Area (Sq Ft)FeatureNUMsq ft
Square footage of interior living or storage housing space situated below ground level.
Oyear_builtYear BuiltFeatureNUMyear
The calendar year in which the primary building construction was completed.
Pyear_renovatedYear RenovatedFeatureNUMyear
The calendar year of the most recent major structural restoration or remodel (0 if never renovated).
Qpostal_codeZIP CodeFeatureNUM
United States Postal Service 5-digit postal routing code representing the local micro-neighborhood.
RlatitudeLatitudeFeatureNUMdegrees
Geographic coordinate indicating the north-south position of the parcel.
SlongitudeLongitudeFeatureNUMdegrees
Geographic coordinate indicating the east-west position of the parcel.
Tnearest_15_living_sqftNearest 15 Neighbors Living Area (Sq Ft)FeatureNUMsq ft
Average interior living square footage of the 15 closest neighboring residential properties.
Unearest_15_lot_sqftNearest 15 Neighbors Lot Size (Sq Ft)FeatureNUMsq ft
Average land lot square footage of the 15 closest neighboring residential properties.
DATASET PREVIEW

Interactive Data Table

Explore rows, feature values, and target labels for House Sales.

Loading dataset...
LOCAL EXECUTION TELEMETRY

TabICLv2 WebGPU Benchmark Results

Complete performance metrics on persisted IID or non-IID evaluation splits, executed 100% locally in the browser sandbox.

Mean Split Latency?Mean wall-clock execution time per persisted evaluation split running 100% locally via WebGPU inside Chrome.Explore Latency Guide →
62.68s9 splits • 8 ensembles/split
Total Evaluation Runtime?Total wall-clock duration across all selected splits, including in-context encoding, WebGPU shader execution, and result aggregation.Explore Split Protocol →
564.16sIncludes warmup & sync
Evaluation Protocol?Precomputed row indices ensure the browser and Python runners evaluate identical out-of-sample observations without leakage.Explore Split Protocol →
3×3 SplitsRepeated IID • 9 total
TabICLv2 Score?Primary task performance score (RMSE) achieved by TabICLv2 across out-of-sample evaluation splits.Explore Metric Formula →
111,524RMSE
Evaluation SplitTrain RowsTest Rows
Split Duration?Wall-clock inference time taken for this evaluation split running 100% locally via WebGPU.Learn more →
R² Score?Coefficient of determination measuring variance explained by the model.Learn more →
MAE?Mean Absolute Error in original target units, robust to outliers.Learn more →
RMSE?Root Mean Squared Error, penalizing large residual deviations.Learn more →
MAPE?Mean Absolute Percentage Error relative to true target values.Learn more →
RMSLE?Root Mean Squared Logarithmic Error, measuring proportional error scales.Learn more →
R1 / F114,4087,20561.09s0.907561,955.13113,519.12
R1 / F214,4097,20463.77s0.883362,449.36131,019.24
R1 / F314,4097,20463.65s0.913756,209.54100,818.54
R2 / F114,4087,20560.91s0.905658,602.19107,041.69
R2 / F214,4097,20463.31s0.892760,981.47120,887.38
R2 / F314,4097,20463.58s0.897861,009.45122,408.74
R3 / F114,4087,20561.21s0.898460,428.37116,400.51
R3 / F214,4097,20463.58s0.900160,379.13111,866.66
R3 / F314,4097,20463.05s0.903060,627.28118,834
Mean ± Std62.68s0.9002 ± 0.008360,293.5115,866.2
UNIFIED BENCHMARK LEADERBOARD

Foundation Models vs Traditional ML Baselines

Side-by-side evaluation of our in-browser Carla engine (WebGPU), open-source tabular foundation models, and OpenML baselines.

RankAlgorithm / ModelModel FamilyRuntime
R² Score?Coefficient of determination (R²) measuring proportion of variance explained.Google ML Guide & Details →
MAE?Mean Absolute Error measuring average prediction error magnitude.Learn more →
RMSE?Root Mean Squared Error penalizing large prediction errors.Learn more →
Reference / Repo
#1Google TabFM v1.0Tabular Foundation ModelPyTorch/Python0.927354,970.998,886.8google-research/tabfm ↗
#2EXAONE TabularTabular Foundation ModelPyTorch/Python0.926455,303.399,519.6LGAI-Research/EXAONE-Tabular ↗
#3TabPFN v3Tabular Foundation ModelPyTorch/Python0.922556,817102,039.4PriorLabs/tabpfn ↗
#4nanotabicl VanillaTabular Foundation ModelPyTorch/Python0.907858,724.6111,387soda-inria/nanotabicl ↗
#5Full TabICLv2 (PyTorch Reference)Tabular Foundation ModelPyTorch/Python0.907658,666.4111,523.7soda-inria/tabicl ↗
#6Streaming nanotabiclTabular Foundation ModelPyTorch/Python0.901759,637.7115,032.2soda-inria/nanotabicl ↗
#7Carla Engine (TabICLv2 WebGPU)Tabular Foundation ModelBrowser0.900260,293.5115,866.2100% In-Browser
STATISTICAL DISTRIBUTIONS

Variable Schema & Summary Distributions

Observed numerical ranges, category cardinalities, missing rates, and sample values across 21,613 rows.

ColVariable NameTypeMissingDistinctSummary Stats / DistributionSample Values
Asale_price_usdTARGETNUM04,028[75000, 7700000] μ=540088.1 σ=367118.7221900538000180000
Bproperty_idNUM021,436[1000102, 9900000190] μ=4580301520.9 σ=2876499023.4712930052064141001925631500400
Csale_dateCAT037220141013T00000020141209T00000020150225T000000
Dbedroom_countNUM013[0, 33] μ=3.4 σ=0.9332
Ebathroom_countNUM030[0, 8] μ=2.1 σ=0.812.251
Fliving_area_sqftNUM01,038[290, 13540] μ=2079.9 σ=918.411802570770
Glot_area_sqftNUM09,782[520, 1651359] μ=15107.0 σ=41419.65650724210000
Hfloor_countNUM06[1, 3.5] μ=1.5 σ=0.5121
Iis_waterfrontNUM02[0, 1] μ=0.0 σ=0.1000
Jview_ratingNUM05[0, 4] μ=0.2 σ=0.8000
Kcondition_ratingNUM05[1, 5] μ=3.4 σ=0.7333
Lconstruction_gradeNUM012[1, 13] μ=7.7 σ=1.2776
Mabove_ground_sqftNUM0946[290, 9410] μ=1788.4 σ=828.111802170770
Nbasement_sqftNUM0306[0, 4820] μ=291.5 σ=442.604000
Oyear_builtNUM0116[1900, 2015] μ=1971.0 σ=29.4195519511933
Pyear_renovatedNUM070[0, 2015] μ=84.4 σ=401.7019910
Qpostal_codeNUM070[98001, 98199] μ=98077.9 σ=53.5981789812598028
RlatitudeNUM05,034[47.1559, 47.7776] μ=47.6 σ=0.147.511247.72100000000000447.7379
SlongitudeNUM0752[-122.519, -121.315] μ=-122.2 σ=0.1-122.257-122.319-122.23299999999999
Tnearest_15_living_sqftNUM0777[399, 6210] μ=1986.6 σ=685.4134016902720
Unearest_15_lot_sqftNUM08,689[651, 871200] μ=12768.5 σ=27303.5565076398062
FREQUENTLY ASKED QUESTIONS

Frequently Asked Questions: House Sales

Common questions regarding the House Sales dataset, machine learning task formulations, and in-browser tabular inference.

What is the King County House Sales dataset used for?

The King County House Sales dataset is a renowned machine learning benchmark used to develop and evaluate Automated Valuation Models (AVMs) and regression algorithms that predict residential home sale prices based on physical attributes, construction quality, and geospatial coordinates.

Can I run zero-server predictions on this dataset in Google Sheets?

Yes, Carla HQ enables zero-server in-context tabular prediction directly in spreadsheets using TabICL foundation models, allowing real estate professionals to forecast property values without configuring Python environments or cloud infrastructure.

What machine learning models perform best on the King County House Sales dataset?

Tree-based gradient boosting algorithms (LightGBM, XGBoost, CatBoost) historically achieve the highest accuracy (R² ~ 0.88–0.90) due to their ability to capture non-linear relationships in spatial coordinates and building grades, while modern tabular foundation models like TabICL offer competitive instant zero-shot performance.

What is the difference between sqft_living and sqft_living15?

The column 'sqft_living' measures the specific property's interior living area, whereas 'sqft_living15' represents the average interior living area of the 15 nearest neighboring properties, acting as a spatial smoothing feature.

LIVE EVALUATION IN GOOGLE SHEETS

Test TabICLv2 on House Sales Yourself

Open the pre-loaded Google Sheet and let Carla configure the target and task for local, zero-cloud tabular machine learning.

Step 1

Launch Carla

Click once to open the spreadsheet and Carla side panel together.

Step 2

Review the Setup

Carla selects the dataset target and task from this page automatically.

Step 3

Evaluate & Predict

Run predictions and compute metrics with zero server uploads.