Skip to main content
The land module provides comprehensive tools for land value modeling through vacant land analysis and hedonic approaches. These functions help separate land value from total property value using various statistical methods.

​
Core Functions

​
run_land_analysis

run_land_analysis(
    sup: SalesUniversePair,
    settings: dict,
    verbose: bool = False
)
Execute land value analysis using vacant and hedonic land models to calculate land allocation ratios. This function processes sales data to determine what percentage of total property value should be attributed to land versus improvements. It runs both vacant land models and hedonic land models, compares their performance, and creates an optimized ensemble allocation.
​
sup
SalesUniversePair
required
Sales and universe data pair containing property information
​
settings
dict
required
Configuration dictionary with modeling instructions and parameters
​
verbose
bool
default:"False"
If True, print detailed progress information during analysis
Process:
  1. Gathers predictions from main, hedonic, and vacant models
  2. Calculates land allocation ratios for each model
  3. Compares model performance using MAPE, R², and RMSE
  4. Optimizes ensemble by iteratively removing worst-performing models
  5. Saves final land analysis results to parquet

​
convolve_land_analysis

convolve_land_analysis(
    sup: SalesUniversePair,
    settings: dict,
    verbose: bool = False
)
Perform spatial convolution analysis on land values to evaluate model performance. This function analyzes land values from vacant land sales, comparing predicted land values per area unit against actual sale prices. It focuses on properties valid for land ratio studies and provides detailed performance metrics.
​
sup
SalesUniversePair
required
Sales and universe data pair
​
settings
dict
required
Settings dictionary with modeling configuration
​
verbose
bool
default:"False"
Enable verbose output for debugging
Outputs:
  • R² (OLS and y=x comparison)
  • Slope of regression line
  • Median sales ratio
  • Coefficient of Dispersion (COD)

​
finalize_land_values

finalize_land_values(
    df_in: pd.DataFrame,
    settings: dict,
    verbose: bool = False
) -> pd.DataFrame
Finalize land values by applying allocation ratios to model predictions.
​
df_in
pd.DataFrame
required
Input DataFrame containing properties to value
​
settings
dict
required
Settings dictionary
​
verbose
bool
default:"False"
Print progress information
​
df_out
pd.DataFrame
GeoDataFrame with columns:
  • model_market_value: Total predicted market value
  • model_land_value: Predicted land value
  • model_impr_value: Predicted improvement value
  • model_land_value_land_{unit}: Land value per area unit
  • model_market_value_land_{unit}: Market value per land area
  • model_market_value_impr_{unit}: Market value per improvement area
Process:
  1. Loads cached land analysis results for each model group
  2. Applies land allocation ratios to calculate land values
  3. Derives improvement values (market value - land value)
  4. Applies quality control checks to land values
  5. Calculates per-area values
  6. Saves final predictions to parquet format

​
Private Helper Functions

​
_run_land_analysis

Internal function that performs the core land analysis workflow for a single model group. Handles model comparison, ensemble optimization, and result persistence.

​
_convolve_land_analysis

Internal function for spatial land value analysis on vacant land sales.

​
_finalize_land_values

Internal function that processes land allocation results and derives final land and improvement values with quality control.

​
Usage Example

from openavmkit.land import run_land_analysis, finalize_land_values
from openavmkit.data import SalesUniversePair

# Run land analysis to calculate allocation ratios
run_land_analysis(sup, settings, verbose=True)

# Apply land values to all properties
df_valued = finalize_land_values(df_universe, settings, verbose=True)

print(df_valued[['key', 'model_market_value', 'model_land_value', 'model_impr_value']].head())

​
Methodology

​
Vacant Land Approach

Models vacant land sales directly to estimate land value per area unit, then applies these rates to improved properties.

​
Hedonic Land Approach

Uses regression models to isolate land characteristics’ contribution to total property value, deriving land allocation ratios statistically.

​
Ensemble Optimization

The module automatically:
  • Compares multiple model approaches
  • Calculates performance metrics (MAPE, R², RMSE)
  • Iteratively removes poorest-performing models
  • Selects optimal ensemble based on:
    • Minimizing negative allocations
    • Minimizing allocations > 100%
    • Maximizing prediction accuracy

Build docs developers (and LLMs) love

Get started for freeTalk to us