R/bt_squad_prediction.R

In global-trade-alert/gtabastiat:

# Roxygen documentation

#' Bastiat, please use the squad classifier to identify relevant text.
#'
#' @return A data frame with the text id and the classification result.
#' @references www.globaltradealert.org
#' @author Johannes Fritz for GTA


# Function infos and parameters  --------------------------------------------

bt_squad_prediction = function(prediction.data.id=NULL,
                               prediction.data.text=NULL,
                               prediction.acting.agency=NULL,
                               train.share=.82,
                               squad.classifier="content/0 core/Bastiat squad classifier.Rdata"
                                 ){

  ## getting squad member predictions
  squad.predictions=data.frame()

  load(squad.classifier)
  rm(classifier, cutoff)
  for(squad.member in bt.squad$member.name){

    load(bt.squad$classifier.location[bt.squad$member.name==squad.member])
    rm(classifier, cutoff)

    #dbg
    # detective=squad.member
    # classifier.location=bt.squad$classifier.location[bt.squad$member.name==squad.member]
    # prediction.word.score=word.score

    squad.predictions=rbind(squad.predictions,
                            bt_detective_prediction(detective=squad.member,
                                                    classifier.location=bt.squad$classifier.location[bt.squad$member.name==squad.member],
                                                    prediction.data.id=prediction.data.id,
                                                    prediction.word.score=word.score,
                                                    prediction.data.text=prediction.data.text,
                                                    prediction.acting.agency=prediction.acting.agency,
                                                    train.share=train.share))

  }




  ## squad variable selection here
  # currently only use prediction probability, not whether it is classified as relevant.
  squad.predictions$relevant=NULL

  squad.predictions=reshape(squad.predictions, idvar = c("bid"), timevar = "detective", direction="wide")

  # Predicting using squad classifier


  predict.bid=squad.predictions$bid
  names(squad.predictions)=gsub(" ","", tolower(names(squad.predictions)))


  ## prediction
  load(squad.classifier)

  predict.data=squad.predictions[,c(classifier$fitLibrary$SL.xgboost_All$object$feature_names)]

  prediction.result= data.frame(text.id=predict.bid,
                                relevant=as.numeric(predict(classifier, predict.data)$pred[,1]>=cutoff),
                                relevance.probability=predict(classifier, predict.data)$pred[,1],
                                stringsAsFactors = F)

  return(prediction.result)

}