Skip to main content

Advertisement

Table 4 Parameters tested using grid-search and 5-fold CV. EFD refers to the “Extended Framework Design”

From: Pipeline design to identify key features and classify the chemotherapy response on lung cancer patients using large-scale genetic data

Pipeline step Parameter options
ANOVA EFD (Partial analysis): percentile = 2% of total # of variables
  EFD (Final analysis): percentile = 10% of total # of variables
  LR penalty = ’l1’
  C = 1
RFE-LR RFE EFD (Partial analysis):
  n_features_to_select = 2% of total # of variables,
  step = 4%
  EFD (Final analysis):
  n_features_to_select = 10% of total # of variables,
  step = 10%
RLR-L1 penalty = ’l1’
  EFD (Partial analysis): C = [100, 500, 1000, 1500, 5000, 10000]
  EFD (Final analysis): C = [100, 500, 1000, 1500, 5000, 10000]
  threshold = 1e−10
Linear SVM C = [0.001, 0.01, 0.1, 1, 10, 100, 1000]
RF n_estimators = [30,47, 75, 119, 189, 299, 475, 753,1194,1892,2999]
KNN n_neighbors = [5, 20, 35, 50]