predict_model: Generate predictions for input data with specified model

In NorskRegnesentral/shapr: Prediction Explanation with Dependence-Aware Shapley Values

Generate predictions for input data with specified model

Description

Performs prediction of response stats::lm(), stats::glm(), ranger::ranger(), mgcv::gam() and xgboost::xgb.train() with binary or continuous response. See details for more information.

NOTE: You should never need to call this function explicitly. It is exported just to be easier accessible for users, see details.

Usage

predict_model(x, newdata, ...)

## Default S3 method:
predict_model(x, newdata, ...)

## S3 method for class 'ar'
predict_model(x, newdata, newreg, horizon, ...)

## S3 method for class 'Arima'
predict_model(
  x,
  newdata,
  newreg,
  horizon,
  explain_idx,
  explain_lags,
  y,
  xreg,
  ...
)

## S3 method for class 'forecast_ARIMA'
predict_model(x, newdata, newreg, horizon, ...)

## S3 method for class 'glm'
predict_model(x, newdata, ...)

## S3 method for class 'lm'
predict_model(x, newdata, ...)

## S3 method for class 'gam'
predict_model(x, newdata, ...)

## S3 method for class 'ranger'
predict_model(x, newdata, ...)

## S3 method for class 'xgb.Booster'
predict_model(x, newdata, ...)

Arguments

x	Model object for the model to be explained.
newdata	A data.frame/data.table with the features to predict from.
...	newreg and horizon parameters used in models passed to ⁠[explain_forecast()]⁠

Details

The following models are currently supported:

• stats::lm()

• stats::glm()

• ranger::ranger()

• mgcv::gam()

• xgboost::xgb.train()

If you have a binary classification model we'll always return the probability prediction for a single class.

If you are explaining a model not supported natively, you need to create the ⁠[predict_model()]⁠ function yourself, and pass it on to as an argument to ⁠[explain()]⁠.

For more details on how to explain such non-supported models (i.e. custom models), see the Advanced usage section of the vignette:
From R: vignette("understanding_shapr", package = "shapr")
Web: https://norskregnesentral.github.io/shapr/articles/understanding_shapr.html#explain-custom-models

Value

Numeric. Vector of size equal to the number of rows in newdata.

Author(s)

Martin Jullum

Examples

# Load example data
data("airquality")
airquality <- airquality[complete.cases(airquality), ]
# Split data into test- and training data
x_train <- head(airquality, -3)
x_explain <- tail(airquality, 3)
# Fit a linear model
model <- lm(Ozone ~ Solar.R + Wind + Temp + Month, data = x_train)

# Predicting for a model with a standardized format
predict_model(x = model, newdata = x_explain)

NorskRegnesentral/shapr documentation built on April 14, 2024, 9:16 a.m.