R/build_forest_predict.R

In IntegratedMRF: Integrated Prediction using Uni-Variate and Multivariate Random Forests

#' Prediction using Random Forest or Multivariate Random Forest
#' 
#' Builds Model of Random Forest or Multivariate Random Forest (when the number of output features > 1) using training samples 
#' and generates the prediction of testing samples using the inferred model.
#'  
#' @param trainX Input Feature matrix of M x N, M is the number of training samples and N is the number of input features 
#' @param trainY Output Response matrix of M x T, M is the number of training samples and T is the number of ouput features
#' @param n_tree Number of trees in the forest, which must be positive integer
#' @param m_feature Number of randomly selected features considered for a split in each regression tree node, which must be positive integer and less than N (number of input features)
#' @param min_leaf Minimum number of samples in the leaf node. If a node has less than or equal to min_leaf samples,
#'  then there will be no splitting in that node and this node will be considered as a leaf node. Valid input is positive integer, which is less than or
#'  equal to M (number of training samples) 
#' @param testX Testing samples of size Q x N, where Q is the number of testing samples and N is the number of features 
#' (Same number of features as training samples)
#' @return Prediction result of the Testing samples 
#' @details
#' Random Forest regression refers to ensembles of regression trees where a set of n_tree un-pruned regression
#' trees are generated based on bootstrap sampling from the original training data. For each node, the optimal
#' feature for node splitting is selected from a random set of m_feature from the total N features. The selection
#' of the feature for node splitting from a random set of features decreases the correlation between different
#' trees and thus the average prediction of multiple regression trees is expected to have lower variance than
#' individual regression trees. Larger m_feature can improve the predictive capability of individual trees but can also
#' increase the correlation between trees and void any gains from averaging multiple predictions. The bootstrap
#' resampling of the data for training each tree also increases the variation between the trees.
#' 
#' In a node with training predictor features (X) and output feature vectors (Y), node splitting is done
#' with the aim of selecting a feature from a random set of m_feature and threshold z to partition the node 
#' into two child nodes, left node (with samples < z) and right node (with samples >=z). In multivariate trees (MRF) 
#' node cost is measured as the sum of squares of the Mahalanobis 
#' distance where as in univariate trees (RF) node cost is measured as the Euclidean distance.
#' 
#' After the Model of the forest is built using training Input features (trainX) and output feature matrix (trainY),
#' the Model is used to generate the prediction of output features (testY) for the testing samples (testX).
#' @examples
#' library(IntegratedMRF)
#' #Input and Output Feature Matrix of random data (created using runif)
#' trainX=matrix(runif(50*100),50,100) 
#' trainY=matrix(runif(50*5),50,5) 
#' n_tree=2
#' m_feature=5
#' min_leaf=5
#' testX=matrix(runif(10*100),10,100) 
#' #Prediction size is 10 x 5, where 10 is the number 
#' #of testing samples and 5 is the number of output features
#' Prediction=build_forest_predict(trainX, trainY, n_tree, m_feature, min_leaf, testX)
#' 
#' @importFrom stats cov
#' @references 
#' [Random Forest] Breiman, Leo. "Random forests." Machine learning 45.1 (2001): 5-32.
#' 
#' [Multivariate Random Forest] Segal, Mark, and Yuanyuan Xiao. "Multivariate random forests." 
#' Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 1.1 (2011): 80-87.
#' @export
#' 
build_forest_predict <- function(trainX, trainY, n_tree, m_feature, min_leaf, testX){
  if (class(n_tree)=="character" || n_tree%%1!=0 || n_tree<1) stop('Number of trees in the forest can not be fractional or negative integer or string')
  if (class(m_feature)=="character" || m_feature%%1!=0 || m_feature<1) stop('Number of randomly selected features considered for a split can not be fractional or negative integer or string')
  if (class(min_leaf)=="character" || min_leaf%%1!=0 || min_leaf<1 || min_leaf>nrow(trainX)) stop('Minimum leaf number can not be fractional or negative integer or string or greater than number of samples')
   
  theta <- function(trainX){trainX}
  results <- bootstrap::bootstrap(1:nrow(trainX),n_tree,theta) 
  b=results$thetastar
  
  Variable_number=ncol(trainY)
  if (Variable_number>1){
    Command=2
  }else if(Variable_number==1){
    Command=1
  } 
  
  Y_HAT=matrix(  0*(1:Variable_number*nrow(testX)),  ncol=Variable_number,   nrow=nrow(testX)  )
  Y_pred=NULL
  
  for (i in 1:n_tree){
    Single_Model=NULL
    X=trainX[ b[ ,i],  ]
    Y=matrix(trainY[ b[ ,i],  ],ncol=Variable_number)
    Inv_Cov_Y = solve(cov(Y)) # calculate the V inverse
    if (Command==1){
      Inv_Cov_Y=matrix(rep(0,4),ncol=2)
    }
    Single_Model=build_single_tree(X, Y, m_feature, min_leaf,Inv_Cov_Y,Command)
    Y_pred=single_tree_prediction(Single_Model,testX,Variable_number)
    for (j in 1:Variable_number){
      Y_HAT[,j]=Y_HAT[,j]+Y_pred[,j]
    }
  }
  Y_HAT=Y_HAT/n_tree
  return(Y_HAT)
}