randomForestPredict_.daph

#-------------------------------------------------------------
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
#
# Modifications 2024 The DAPHNE Consortium.
#
#-------------------------------------------------------------

# This script has been manually translated from Apache SystemDS (https://github.com/apache/systemds).
# Original file: scripts/builtin/decisionTree.dml @ 4c97c1c376727f552a80c60217648c37d76fcb4e.

# This script implements random forest prediction for recoded and binned
# categorical and numerical input features.
#
# INPUT:
# ------------------------------------------------------------------------------
# X               Feature matrix in recoded/binned representation
# y               Label matrix in recoded/binned representation,
#                 optional for accuracy evaluation
# ctypes          Row-Vector of column types [1 scale/ordinal, 2 categorical]
# M               Matrix M holding the learned trees (one tree per row),
#                 see randomForest() for the detailed tree representation.
# verbose         Flag indicating verbose debug output
# ------------------------------------------------------------------------------
#
# OUTPUT:
# ------------------------------------------------------------------------------
# yhat            Label vector of predictions
# ------------------------------------------------------------------------------

import "decisionTreePredict_.daph";
import "lmPredictStats_.daph";

# TODO Support optional parameters with defaults (see #548).
def randomForestPredict(X:matrix<f64>, y:matrix<f64> /*= fill(0.0,0,0)*/,
    ctypes:matrix<f64>, M:matrix<f64>, verbose:bool /*= false*/) -> matrix<f64>
{
  t1 = now();
  classify = as.scalar(ctypes[0,ncol(X)]) == 2;
  yExists = (nrow(X)==nrow(y));

  if(verbose) {
    print("randomForestPredict: called for batch of "+nrow(X)+" rows, model of "
      +nrow(M)+" trees, and with labels-provided "+yExists+".");
  }

  # scoring of num_tree decision trees
  Ytmp = fill(0.0, nrow(M), nrow(X));
  # TODO Support parfor-loops (see #515).
  for(i in 1:nrow(M)) {
    if( verbose )
      print("randomForestPredict: start scoring tree "+i+"/"+nrow(M)+".");

    # step 1: sample features (consistent with training)
    I2 = M[i - 1, 0:ncol(X)];
    Xi = X[[, t(I2)]];

    # step 2: score decision tree
    t2 = now();
    
    ret = decisionTreePredict_.decisionTreePredict(Xi, ctypes, M[i - 1,ncol(X):ncol(M)], "TT", false);

    #decisionTreePredict(X:matrix<f64>, y:matrix<f64> /*= fill(0.0,0,0)*/,
    #ctypes:matrix<f64>, M:matrix<f64>, strategy:str /*="TT"*/, verbose:bool /*= false*/)

    Ytmp[i - 1,0:nrow(ret)] = t(ret);
    if( verbose )
      print("-- ["+i+"] scored decision tree in "+as.f64(now()-t2)/1e9+" seconds.");
  }

  # label aggregation (majority voting / average)
  yhat = fill(0.0, nrow(X), 1);
  if( classify ) {
    # TODO Support parfor-loops (see #515).
    for(i in 1:nrow(X))
      yhat[i - 1,0] = as.f64(idxMax(t(ctable(Ytmp[,i - 1] - 1, fill(0.0, nrow(Ytmp), 1))), 0) + 1);
  }
  else {
    yhat = t(sum(Ytmp, 1)/nrow(M));
  }

  # summary statistics
  if( yExists && verbose ) {
    if( classify )
      print("Accuracy (%): " + (sum(yhat == y) / nrow(y) * 100));
    else
      lmPredictStats_.lmPredictStats(yhat, y, false, verbose);
  }

  if(verbose) {
    print("randomForestPredict: scored batch of "+nrow(X)+" rows with "+nrow(M)+" trees in "+as.f64(now()-t1)/1e9+" seconds.");
  }

  return yhat;
}