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

omisvm() has reasonable performance

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.9515
Code
  print(mzoe)
Output
  [1] 0.2
Code
  print(mae)
Output
  [1] 0.2
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4  5
     1  9  1  0  0  0
     2  1 32  1  0  0
     3  0  7 12  4  0
     4  0  0  2 16  2
     5  0  0  0  2 11

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.9333
Code
  print(mzoe)
Output
  [1] 0.28
Code
  print(mae)
Output
  [1] 0.28
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4  5
     1 11  6  0  0  0
     2  2 26  3  0  0
     3  0  9 14  2  0
     4  0  0  5  9  0
     5  0  0  0  1 12

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.8741
Code
  print(mzoe)
Output
  [1] 0.32
Code
  print(mae)
Output
  [1] 0.32
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4
     1  6  4  0  0
     2  0 33  1  0
     3  0  4  9 10
     4  0  0  0 20
     5  0  0  0 13

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.8392
Code
  print(mzoe)
Output
  [1] 0.48
Code
  print(mae)
Output
  [1] 0.5
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4
     1  4 13  0  0
     2  1 27  1  2
     3  0  8  8  9
     4  0  0  1 13
     5  0  0  0 13

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.8753
Code
  print(mzoe)
Output
  [1] 0.32
Code
  print(mae)
Output
  [1] 0.32
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4
     1  6  4  0  0
     2  0 33  1  0
     3  0  5  9  9
     4  0  0  0 20
     5  0  0  0 13

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.8426
Code
  print(mzoe)
Output
  [1] 0.48
Code
  print(mae)
Output
  [1] 0.5
Code
  print(table(true, pred))
Output
      pred
  true  1  2  3  4
     1  4 13  0  0
     2  2 26  1  2
     3  0  9  9  7
     4  0  0  1 13
     5  0  0  0 13

`omisvm()` value returns make sense

Code
  models <- list(xy = .run_omisvm(df1, weights = NULL), formula = omisvm(mi(
    bag_label, bag_name) ~ V1 + V2, data = df1, weights = NULL), mi_df = omisvm(
    as_mi_df(df1, instance_label = NULL)), `no-scale` = .run_omisvm(df1, weights = NULL,
    control = list(scale = FALSE))) %>% suppressWarnings() %>% suppressMessages()
  print(lapply(models, names))
Output
  $xy
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "h"          "s"          "kernel"     "repr_inst"  "x_scale"

  $formula
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "h"          "s"          "kernel"     "repr_inst"  "x_scale"   
  [11] "formula"    "bag_name"

  $mi_df
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "h"          "s"          "kernel"     "repr_inst"  "x_scale"   
  [11] "bag_name"

  $`no-scale`
  [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
  [6] "h"          "s"          "kernel"     "repr_inst"

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

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    h: 1 
    s: 2 
    scale: TRUE 
    weights: FALSE

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


  $formula
  An misvm object called with omisvm.formula

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    h: 1 
    s: 2 
    scale: TRUE 
    weights: FALSE

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


  $mi_df
  An misvm object called with omisvm.mi_df

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    h: 1 
    s: 2 
    scale: TRUE 
    weights: FALSE

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


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

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    h: 1 
    s: 2 
    scale: FALSE 
    weights: FALSE

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

Ordering of data doesn't change `omisvm()` results

Code
  with(df1_test, suppressWarnings({
    pred <- predict(mdl2, df1_test, type = "raw")$.pred
    pROC::auc(classify_bags(bag_label, bag_name), classify_bags(pred, bag_name))
  }))
Message <simpleMessage>
  Setting levels: control = 1, case = 2
  Setting direction: controls < cases
Output
  Area under the curve: 0.9335