R/model_structured_data_classifier.R
model_structured_data_classifier.RdAutoKeras structured data classification class.
To `fit`, `evaluate` or `predict`, format inputs as:
x : character or array. If the data is from a csv file, it should be a character specifying the path of the csv file of the training data.
y : character or array. If the data is from a csv file, it should be a character, which is the name of the target column. Otherwise, It can be raw labels, one-hot encoded if more than two classes, or binary encoded for binary classification.
model_structured_data_classifier( column_names = NULL, column_types = NULL, num_classes = NULL, multi_label = FALSE, loss = NULL, metrics = NULL, name = "structured_data_classifier", max_trials = 100, directory = tempdir(), objective = "val_accuracy", overwrite = TRUE, seed = runif(1, 0, 1e+07) )
| column_names | : A list of characters specifying the names of the columns. The length of the list should be equal to the number of columns of the data excluding the target column. Defaults to `NULL`. If `NULL`, it will obtained from the header of the csv file or the `data.frame`. |
|---|---|
| column_types | : A list of characters. The names are the column names. The values should either be 'numerical' or 'categorical', indicating the type of that column. Defaults to `NULL`. If not `NULL`, the `column_names` need to be specified. If `NULL`, it will be inferred from the data. |
| num_classes | : numeric. Defaults to `NULL`. If `NULL`, it will infer from the data. |
| multi_label | : logical. Defaults to `FALSE`. |
| loss | : A Keras loss function. Defaults to use "binary_crossentropy" or "categorical_crossentropy" based on the number of classes. |
| metrics | : A list of Keras metrics. Defaults to use "accuracy". |
| name | : character. The name of the AutoModel. Defaults to "structured_data_classifier". |
| max_trials | : numeric. The maximum number of different Keras Models to try. The search may finish before reaching the `max_trials`. Defaults to `100`. |
| directory | : character. The path to a directory for storing the search outputs. Defaults to `tempdir()`, which would create a folder with the name of the AutoModel in the current directory. |
| objective | : character. Name of model metric to minimize or maximize. Defaults to "val_accuracy". |
| overwrite | : logical. Defaults to `TRUE`. If `FALSE`, reloads an existing project of the same name if one is found. Otherwise, overwrites the project. |
| seed | : numeric. Random seed. Defaults to `runif(1, 0, 10e6)`. |
A non-trained structured data classifier AutokerasModel.
Important: The object returned by this function behaves like an R6 object, i.e., within function calls with this object as parameter, it is most likely that the object will be modified. Therefore it is not necessary to assign the result of the functions to the same object.
if (FALSE) { library("magrittr") # use the iris dataset as an example set.seed(8818) # balanced sample 80% for training train_idxs <- unlist(by(seq_len(nrow(iris)), iris$Species, function(x) { sample(x, length(x) * .8) })) train_data <- iris[train_idxs, ] test_data <- iris[-train_idxs, ] colnames(iris) # Species will be the interest column to predict train_file <- paste0(tempdir(), "/iris_train.csv") write.csv(train_data, train_file, row.names = FALSE) # file to predict, cant have the response "Species" column test_file_to_predict <- paste0(tempdir(), "/iris_test_2_pred.csv") write.csv(test_data[, -5], test_file_to_predict, row.names = FALSE) test_file_to_eval <- paste0(tempdir(), "/iris_test_2_eval.csv") write.csv(test_data, test_file_to_eval, row.names = FALSE) library("autokeras") # Initialize the structured data classifier clf <- model_structured_data_classifier(max_trials = 10) %>% # It tries 10 different models fit(train_file, "Species") # Feed the structured data classifier with training data # If you want to use own valitadion data do: clf <- model_structured_data_classifier(max_trials = 10) %>% fit( train_file, "Species", validation_data = list(test_file_to_eval, "Species") ) # Predict with the best model (predicted_y <- clf %>% predict(test_file_to_predict)) # Evaluate the best model with testing data clf %>% evaluate(test_file_to_eval, "Species") # Get the best trained Keras model, to work with the keras R library export_model(clf) }