TwoStageModel

The core model for predicting spatial variations in ML model accuracy.

from geoequity import TwoStageModel

model = TwoStageModel(spline=7, lam=0.5, resolution=[30, 30])

Constructor

Parameter	Type	Default	Description
spline	int	7	Number of spline bases for GAM
lam	float	0.5	GAM regularization parameter
resolution	list	[30, 30]	Grid resolution for SVM

Methods

fit

def fit(self, df, model_col, target_col='observed', 
        density_col='density', lon_col='longitude', lat_col='latitude'):

Train the two-stage model on test data.

Parameters:

Name	Type	Description
df	DataFrame	Test data with predictions and ground truth
model_col	str	Column name containing model predictions
target_col	str	Column name containing ground truth values
density_col	str	Column name containing density values
lon_col	str	Column name for longitude
lat_col	str	Column name for latitude

predict

def predict(self, longitude, latitude, density):

Predict accuracy for given coordinates and density.

Parameters:

Name	Type	Description
longitude	float or array	Longitude(s) to predict
latitude	float or array	Latitude(s) to predict
density	float or array	Density value(s)

Returns: Predicted R² score(s)

diagnose

def diagnose(self, save_dir='diagnostics/', show=True):

Generate diagnostic plots and summary report.

Parameters:

Name	Type	Description
save_dir	str	Directory to save outputs
show	bool	Whether to display plots

Outputs: - stage1_gam.png: GAM partial dependence plots - stage2_svm.png: Spatial residual map - diagnosis.txt: Summary statistics