svor_exc.md
In mildsvm: Multiple-Instance Learning with Support Vector Machines

svor_exc() internal functions work on simple examples

Code
  table(y, y_pred)
Output
     y_pred
  y     1   2   3   4   5
    1 103   0   0   0   0
    2  13   8   1   0   0
    3   1   2   8   1   0
    4   0   0   1   7   1
    5   0   0   0   2   2
Code
  pROC::multiclass.roc(response = y, predictor = f) %>% suppressMessages()
Output

  Call:
  multiclass.roc.default(response = y, predictor = f)

  Data: f with 5 levels of y: 1, 2, 3, 4, 5.
  Multi-class area under the curve: 0.9833
Code
  mzoe <- mean(y != y_pred)
  mae <- mean(y - y_pred)
  mzoe
Output
  [1] 0.1466667
Code
  mae
Output
  [1] 0.1133333

svor_exc() has reasonable performance

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.9509
Code
  print(mzoe)
Output
  [1] 0.1066667
Code
  print(mae)
Output
  [1] 0.1066667

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.9491
Code
  print(mzoe)
Output
  [1] 0.1235294
Code
  print(mae)
Output
  [1] 0.1235294

`svor_exc()` value returns make sense

Code
  models <- list(xy = svor_exc(x = df1[, 2:6], y = df1$y, weights = NULL),
  formula = svor_exc(y ~ V1 + V2, data = df1, weights = NULL), mi_df = svor_exc(
    as_mi_df(df2, bag_label = "y", instance_label = NULL)), `no-scale` = svor_exc(
    x = df1[, 2:6], y = df1$y, weights = NULL, control = list(scale = FALSE))) %>%
    suppressWarnings() %>% suppressMessages()
  print(lapply(models, names))
Output
  $xy
  [1] "smo_fit"   "call_type" "x"         "features"  "levels"    "cost"     
  [7] "kernel"    "n_step"    "x_scale"

  $formula
   [1] "smo_fit"   "call_type" "x"         "features"  "levels"    "cost"     
   [7] "kernel"    "n_step"    "x_scale"   "formula"

  $mi_df
   [1] "smo_fit"   "call_type" "x"         "features"  "levels"    "cost"     
   [7] "kernel"    "n_step"    "x_scale"   "bag_name"

  $`no-scale`
  [1] "smo_fit"   "call_type" "x"         "features"  "levels"    "cost"     
  [7] "kernel"    "n_step"

Code
  print(models)
Output
  $xy
  An svor_exc object called with svor_exc.default

  Parameters: 
    method: smo 
    kernel: linear  
    cost: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:5] "1" "2" "3" "4" "5"
    Features: chr [1:5] "V1" "V2" "V3" "V4" "V5"
    Number of iterations: 500


  $formula
  An svor_exc object called with svor_exc.formula

  Parameters: 
    method: smo 
    kernel: linear  
    cost: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:5] "1" "2" "3" "4" "5"
    Features: chr [1:2] "V1" "V2"
    Number of iterations: 146


  $mi_df
  An svor_exc object called with svor_exc.mi_df

  Parameters: 
    method: smo 
    kernel: linear  
    cost: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:5] "1" "2" "3" "4" "5"
    Features: chr [1:5] "V1" "V2" "V3" "V4" "V5"
    Number of iterations: 500


  $`no-scale`
  An svor_exc object called with svor_exc.default

  Parameters: 
    method: smo 
    kernel: linear  
    cost: 1 
    scale: FALSE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:5] "1" "2" "3" "4" "5"
    Features: chr [1:5] "V1" "V2" "V3" "V4" "V5"
    Number of iterations: 500