BioSeqClass (version 1.30.0)

classify: Classification with Specific Features and Cross-Validation

Description

Classification with selected features and cross-validation. It supports 10 classification algorithms, feature selection by Weka, cross-validation and leave-one-out test.

Usage

classify(data,classifyMethod="libsvm",cv=10, features, evaluator, search, n=200,         svm.kernel="linear",svm.scale=FALSE,  svm.path, svm.options="-t 0", knn.k=1, nnet.size=2, nnet.rang=0.7, nnet.decay=0, nnet.maxit=100)

Arguments

data

a data frame including the feature matrix and class label. The last column is a vector of class label comprising of "-1" or "+1"; Other columns are features.

classifyMethod

a string for the classification method. This must be one of the strings "libsvm", "svmlight", "NaiveBayes", "randomForest", "knn", "tree", "nnet", "rpart", "ctree", "ctreelibsvm", "bagging".

cv

an integer for the time of cross validation, or a string "leave\_one\_out" for the jackknife test.

features

an integer vector for the index of interested columns in data, which will be used as features for build classification model.

evaluator

a string for the feature selection method used by WEKA. This must be one of the strings "CfsSubsetEval", "ChiSquaredAttributeEval", "InfoGainAttributeEval", or "SVMAttributeEval".

search

a string for the search method used by WEKA. This must be one of the strings "BestFirst" or "Ranker".

n

an integer for the number of selected features.

svm.kernel

a string for kernel function of SVM.

svm.scale

a logical vector indicating the variables to be scaled.

svm.path

a character for path to SVMlight binaries (required, if path is unknown by the OS).

svm.options

Optional parameters to SVMlight. For further details see: "How to use" on http://svmlight.joachims.org/. (e.g.: "-t 2 -g 0.1"))

nnet.size

number of units in the hidden layer. Can be zero if there are skip-layer units.

nnet.rang

Initial random weights on [-rang, rang]. Value about 0.5 unless the inputs are large, in which case it should be chosen so that rang * max(|x|) is about 1.

nnet.decay

parameter for weight decay.

nnet.maxit

maximum number of iterations.

knn.k

number of neighbours considered in function classifyModelKNN.

Details

classify employ feature selction method in Weka and diverse classification model in other R packages to perfrom classification. "Cross Validation" is controlled by parameter "cv"; "Feature Selection" is controlled by parameter "features", "evaluator", "search", and "n"; "Classification Model Building" is controlled by parameter "classifyMethod".

Parameter "evaluator" supportes three feature selection methods provided by WEKA: "CfsSubsetEval": Evaluate the worth of a subset of attributes by considering the individual predictive ability of each feature along with the degree of redundancy between them. "ChiSquaredAttributeEval": Evaluate the worth of an attribute by computing the value of the chi-squared statistic with respect to the class. "InfoGainAttributeEval": Evaluate attributes individually by measuring information gain with respect to the class. "SVMAttributeEval": Evaluate the worth of an attribute by using an SVM classifier. Attributes are ranked by the square of the weight assigned by the SVM. Attribute selection for multiclass problems is handled by ranking attributes for each class seperately using a one-vs-all method and then "dealing" from the top of each pile to give a final ranking. Parameter "search" supportes three feature subset search methods provided by WEKA: "BestFirst": Searches the space of attribute subsets by greedy hillclimbing augmented with a backtracking facility. Setting the number of consecutive non-improving nodes allowed controls the level of backtracking done. Best first may start with the empty set of attributes and search forward, or start with the full set of attributes and search backward, or start at any point and search in both directions (by considering all possible single attribute additions and deletions at a given point). "Ranker": Ranks attributes by their individual evaluations.

Parameter "classifyMethod" supports multiple classification model: "libsvm": Employ classifyModelLIBSVM to perform Support Vecotr Machine by LibSVM. Package "e1071" is required. "svmlight": Employ classifyModelSVMLIGHT to Support Vecotr Machine by SVMLight. Package "klaR" is required. "NaiveBayes": Employ classifyModelNB to perform Naive Bayes classification. Package "klaR" is required. "randomForest": Employ classifyModelRF to perform random forest classification. Package "randomForest" is required. "knn": Employ classifyModelKNN to perform k Nearest Neighbor algorithm. Package "class" is required. "tree": Employ classifyModelTree to perform tree classification. Package "tree" is required. "nnet": Employ classifyModelNNET to perform neural net algorithm. Bundle "VR" is required. "rpart": Employ classifyModelRPART to perform Recursive Partitioning and Regression Trees. Package "rpart" is required. "ctree": Employ classifyModelCTREE to perform Conditional Inference Trees. Package "party" is required. "ctreelibsvm": Employ classifyModelCTREELIBSVM to combine Conditional Inference Trees and Support Vecotr Machine for classification. For each node in the tree, one SVM model will be constructed using train data in this node. Test data will be firstly classified to one node of the tree, and then use corresponding SVM to do classification. Package "party" and "e1071" is required. "bagging": Employ classifyModelBAG to perform bagging for classification trees. Package "ipred" is required.

Examples

  ## read positive/negative sequence from files.
  tmpfile1 = file.path(path.package("BioSeqClass"), "example", "acetylation_K.pos40.pep")
  tmpfile2 = file.path(path.package("BioSeqClass"), "example", "acetylation_K.neg40.pep")
  posSeq = as.matrix(read.csv(tmpfile1,header=FALSE,sep="\t",row.names=1))[,1]
  negSeq = as.matrix(read.csv(tmpfile2,header=FALSE,sep="\t",row.names=1))[,1]
  seq=c(posSeq,negSeq)
  classLable=c(rep("+1",length(posSeq)),rep("-1",length(negSeq)) ) 
  data = data.frame(featureBinary(seq),classLable)
  
  ## Use LibSVM and 5-cross-validation to classify.
  LIBSVM_CV5 = classify(data,classifyMethod="libsvm",cv=5,
                 svm.kernel="linear",svm.scale=FALSE)
  ## Features selection is done by envoking "CfsSubsetEval" method in WEKA.               
  FS_LIBSVM_CV5 = classify(data,classifyMethod="libsvm",cv=5,evaluator="CfsSubsetEval",
                 search="BestFirst",svm.kernel="linear",svm.scale=FALSE)    
  
  if(interactive()){
  
    KNN_CV5 = classify(data,classifyMethod="knn",cv=5,knn.k=1)  
    
    RF_CV5 = classify(data,classifyMethod="randomForest",cv=5)
    
    TREE_CV5 = classify(data,classifyMethod="tree",cv=5)
    
    NNET_CV5 = classify(data,classifyMethod="nnet",cv=5)
    
    RPART_CV5 = classify(data,classifyMethod="rpart",cv=5,evaluator="")
    
    CTREE_CV5 = classify(data,classifyMethod="ctree",cv=5,evaluator="")  
    
    BAG_CV5 = classify(data,classifyMethod="bagging",cv=5,evaluator="")  
         
  }