getMLmodel is a function for retrain a deep model with a prefixed configuration

Usage

getMLmodel.withRetrain(
  dataset = NULL,
  depVar = "Cab",
  model = "CNN",
  optimizer = "adam",
  n.times = NULL,
  n.neurons = 128,
  n.layers = 4,
  batch.size = 125,
  n.epochs = 100,
  save.model = T,
  path.model = NULL,
  prop.split = c(0.8, 0.2),
  data.trans = "preProcess",
  method.preProcess = "Normalize",
  depVar.trans = FALSE
)

Arguments

dataset: a dataframe
depVar: name of the variable to predict
model: a ML model. options are: 'CNN','Hidden-layers',
optimizer: the optimizer for the model. options are: 'adam','adadelta','adagrad', 'adamax', 'nadam', 'msprop', 'sgd'
n.times: number of times to repaet the model. By default is 1
n.neurons: number of neurons to use by the the model. By default is 128 (only for hidden layer model is implemented)
n.layers: number of hidden-layer to use by the the model. By default is 4 (only for hidden layer model is implemented))
batch.size: batch size used for each epoch. By default is 125
n.epochs: number of epoch. By default is 100
save.model: a boolean variable for saving ML model, options are: TRUE or FALSE. if TRUE, please use path.model to give a folder for the model
path.model: a path for saving the models. By default path.model ='Models'
prop.split: a vector with proportion for spliting the dataset. prop.split =c(0.8,02) will be used as default.
data.trans: a data.transformation method, options are: 'PCA','preProcess',
method.preProcess: the data.transformation method for preProcess, data.transformation are: 'Normalize', 'YeoJohnson','BoxCox', Standarize', 'Center','Scale', and 'PCA'
depVar.trans: a boolean variable for applying data transformation in Y variable, options are: TRUE or FALSE.

Value

a list with models and plots