training_config.yml

sensor_combinations:
  - acc-X
  - acc-Y
  - acc-Z
  - gyr-X
  - gyr-Y
  - gyr-Z

segmentation:
  window: !!float 1.0
  overlap: !!float 0.75

# Based on the NIOSH Equation
data_labels:
    Crate-W2: !!int 0
    Crate-W5: !!int 0
    Crate-W10: !!int 0
    Crate-W15: !!int 0
    Crate-W20: !!int 1
    Crate-W30: !!int 1
    CardboardBox-W2: !!int 0
    CardboardBox-W5: !!int 0
    CardboardBox-W10: !!int 0
    CardboardBox-W15: !!int 0
    CardboardBox-W20: !!int 1
    CardboardBox-W30: !!int 1

# Model parameters for training - Based on lib/models.py
model_params:
  LogisticRegression:
    class_weight: balanced
    max_iter: !!int 5000
    n_jobs: !!int 4
    tol: !!float 0.0001
  DecisionTreeClassifier:
    class_weight: balanced
    max_depth: !!int 50
    min_samples_leaf: !!int 20
    min_samples_split: !!int 20
  KNeighborsClassifier:
    n_neighbors: !!int 10
    weights: uniform
  SVC:
    class_weight: balanced
    kernel: poly
    tol: !!float 1e-07
  BaggingClassifier:
    n_estimators: !!int 100
  RandomForestClassifier:
    class_weight: balanced
    max_depth: !!int 50
    min_samples_leaf: !!int 1
    min_samples_split: !!int 100
    n_estimators: !!int 100
  MLPClassifier:
    hidden_layer_sizes: !!python/tuple [50, 20]
    max_iter: !!int 500
  AdaBoostClassifier:
    base_estimator: !!python/BaggingClassifier {n_estimators: 100}
    n_estimators: 500
    learning_rate: 0.001

# Sensor(s) for training based on location
tr_sensor_locations:
  - L1
  - L2
  - L3

# Trained model save location
model_save_directory: trained_models