R/evaluate.ppred_model.R

In processpredictR: Process Prediction

#' Generic evaluate function for ppred_model class.
#' @param object [`ppred_model`] (default [`NULL`]): ProcessTransformer model of class [`ppred_model`].
#' @param test_data [`ppred_examples_df`] (default [`NULL`]): preprocessed test data.
#' @param ... Additional arguments
#'
#' @importFrom keras evaluate
#' @export
keras::evaluate

#' @export
evaluate.ppred_model <- function(object, test_data, ...) {


  if(object$task == "remaining_trace_s2s") {
    evaluate_s2s(object, test_data, ...)
  } else {

    # if test_data is a preprocessed test dataset (before tokenize)
    if (any((test_data %>% class) == "ppred_examples_df")) {
      tokens_test <- test_data %>% tokenize()
    }

    # if test_data is already tokenized
    else if (any((test_data %>% class) == "ppred_examples_tokens")) {
      tokens_test <- test_data
    }

    # x_test ------------------------------------------------------------------
    # tokenized traces
    x_tokens_test <- tokens_test$token_x %>% keras::pad_sequences(maxlen = object$max_case_length, value = 0)

    # extra numeric and categorical features (if present)
    x_numeric_features <- tokens_test$numeric_features
    x_categorical_features <- tokens_test$categorical_features
    x_test_list <- list(x_tokens_test)
    if (!is.null(x_numeric_features)) x_test_list <- x_test_list %>% append(list(x_numeric_features))
    if (!is.null(x_categorical_features)) x_test_list <- x_test_list %>% append(list(x_categorical_features))

    # y_test ------------------------------------------------------------------
    y_tokens_test <- tokens_test$token_y

    # NEXT_TIME & REMAINING_TIME
    if (object$task %in% c("next_time", "remaining_time")) {

      metrics <- list(...)
      if(length(metrics) == 0) {

        #######################    #######################    #######################    #######################
        y_tokens_test <- y_tokens_test / object$sd_time

        # should be based solely on the metrics from compile()
        results <- keras::evaluate(object$model, x_test_list, y_tokens_test, ...) #%>% keras::k_argmax(axis = -1)
        return(results)

        # # original
        # mean <- object$y_normalize_layer$mean %>% as.double()
        # variance <- object$y_normalize_layer$variance %>% as.double()
        # y_tokens_test <- keras::layer_normalization(y_tokens_test, mean = mean, variance = variance)

        #######################    #######################    #######################    #######################
        # normalize_y <- keras::layer_normalization()
        # normalize_y %>% adapt(y_tokens_test)
        # y_tokens_test <- normalize_y(y_tokens_test)
      }

      else {
        results <- predict(object, test_data, output = "append")

        # for(i in 1:length(metrics)) {
        #   metrics[[i]](y_tokens_test, results)
        # }

        y_pred <- NULL
        y_var <- if_else(object$task == "next_time", "time_till_next_activity", "remaining_time")
        results %>% summarize(
          across(y_var,
                 .fns = list(...),
                 y_pred,
                 .names = "metric_{.fn}"))

        # tmppred %>% summarize(across(time_till_next_activity,
        #                              .fns = list(mae = Metrics::mae, rmse = Metrics::rmse),
        #                              y_pred,
        #                              .names = "metric_{.fn}") )

      }
    }

    else {
      results <- keras::evaluate(object$model, x_test_list, y_tokens_test, ...) #%>% keras::k_argmax(axis = -1)
      return(results)
    }
  }
}


#### PREDICT ######
# Metrics::rmse()