R/coverageFun.R

In viadee/anchorsOnR: AnchorsOnR: High-Precision Model-Agnostic Explanations in R

calculate.coverage <-
  function(instance,
           features,
           bins,
           coverage_perturbations) {
    featureVec <- as.data.frame(unclass(instance[, features]))
    colnames(featureVec) <- features
    reducedPerturbations <-
      as.data.frame(unclass(coverage_perturbations[, features]))
    colnames(reducedPerturbations) = features

    for (i in 1:ncol(reducedPerturbations)) {
      featureVec[i] = provideBin(featureVec[i], bins[[features[i]]])
    }

    matchingRows = nrow(suppressMessages(plyr::match_df(reducedPerturbations, featureVec)))
    coverage = matchingRows / nrow(reducedPerturbations)

    return(coverage)
  }