diabetes-pytorch

Please see notebooks/PR-Diabetes-Results.ipynb for full results!

Excerpts from Results:

From UCI ML Repository
"Diabetes 130-US hospitals for years 1999-2008 Data Set"

Goal: build model(s) to predict which patients will be re-hospitalized within 30 days

Evaluate: using AUROC

Notes:

• 'encounter_id' - unique admissions
• ignore 'patient_nbr' - treat all encounters independent
• 'readmitted' - treat 'NO' as '>30'
• Attributes: 55
• Samples: >100k
• Features: numerical, categorical

Results

Metric	Train	Val	Test
Accuracy	0.6164	0.6074	0.6260
AUROC	0.6570	0.6428	0.6650
Loss	0.6584	0.6653	0.6579

Model

Mixed-input deep neural network with categorical embeddings

Network(
  (embeddings): ModuleList(
    (0): Embedding(6, 3)
    (1): Embedding(2, 1)
    (2): Embedding(4, 2)
    (3): Embedding(22, 11)
    (4): Embedding(15, 8)
    (5): Embedding(9, 5)
    (6): Embedding(10, 5)
    (7): Embedding(10, 5)
    (8): Embedding(4, 2)
    (9): Embedding(4, 2)
    (10): Embedding(4, 2)
    (11): Embedding(4, 2)
    (12): Embedding(4, 2)
    (13): Embedding(4, 2)
    (14): Embedding(4, 2)
    (15): Embedding(4, 2)
    (16): Embedding(2, 1)
    (17): Embedding(2, 1)
  )
  (dropout_emb): Dropout(p=0.5)
  (bn_continuous): BatchNorm1d(9, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (fc_layers): ModuleList(
    (0): FCUnit(
      (linear): Linear(in_features=67, out_features=512, bias=True)
      (batchnorm): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (1): FCUnit(
      (linear): Linear(in_features=512, out_features=256, bias=True)
      (batchnorm): BatchNorm1d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (2): FCUnit(
      (linear): Linear(in_features=256, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (3): FCUnit(
      (linear): Linear(in_features=64, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (4): FCUnit(
      (linear): Linear(in_features=64, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (5): FCUnit(
      (linear): Linear(in_features=64, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (6): FCUnit(
      (linear): Linear(in_features=64, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (7): FCUnit(
      (linear): Linear(in_features=64, out_features=64, bias=True)
      (batchnorm): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
    (8): FCUnit(
      (linear): Linear(in_features=64, out_features=32, bias=True)
      (batchnorm): BatchNorm1d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (dropout): Dropout(p=0.65)
    )
  )
  (output_linear): Linear(in_features=32, out_features=1, bias=True)
  (sigmoid): Sigmoid()
)