Skip to main content

Table 4 The best parameters found as a result of the parameter tuning phase for the classification models

From: Statistical representation models for mutation information within genomic data

Algorithm Parameter Value Data Rep.
KNN k 50 Binary
   10 c-score, tf-idf, tf-rf,
    bm25-tf-idf, bm25-tf-rf
SVM-poly Polynomial degree 3 binary, tf-idf, bm25-tf-idf
   2 c-score, tf-rf, bm25-tf-rf
SVM-rbf Gamma 10−4 All
  Cost 103 All
SVM-linear Gamma 10−4 All
  Cost 102 All
Perceptron Optimization function SGD binary
   Adam c-score, tf-idf, tf-rf,
    bm25-tf-idf, bm25-tf-rf
  Activation function tanh binary
   ReLU c-score, tf-idf, tf-rf,
    bm25-tf-idf, bm25-tf-rf
  Hidden layer size 100 All
  The maximum number of iterations 200 binary
   300 c-score, tf-idf, tf-rf,
    bm25-tf-idf, bm25-tf-rf
Feed Forward NN Optimization function Adam All
  Activation function ReLU All
  The number of layers 4 All
  Dropout rate 0.25 All
  The number of nodes in the first layer 8192 All