pyROGER.roger

class pyROGER.roger.RogerModel(x_dataset, y_dataset, ml_models=[KNeighborsClassifier(n_neighbors=63), RandomForestClassifier(max_depth=2, random_state=0), SVC(probability=True)], train_indices=None, test_indices=None, train_percentage=0.75, labels=['CL', 'RIN', 'BS', 'IN', 'ITL'], split_seed=None, comments=None)

Bases: object

Main class of pyROGER. This object contains all the information of the classification.

x_dataset

Numpy array containing the features used for the classification. Usually these will be the relative velocity normalized to the velocity dispersion and the normalized radii.

Type:

np.array

y_dataset

Numpy array containing the real classes of the galaxies. Usually these are cluster, recent infalling, backsplash, infalling and interloper galaxies.

Type:

np.array

ml_models

List with the machine learning methods that will be used. The default methods are KNN, random forest and SVM.

Type:

list

train_percentage

Float between 0 and 1 representing the percentage of observations that will be used for training.

Type:

float

split_seed

Random seed used for randomly splitting the data. Default = None

Type:

Int.

comments

String adding information for the trained model.

Type:

str

split()

Split the data into training and testing set. It returns the indices of both sets. It will run when instantiating a RogerModel object.

train()

Train all the methods contained in ml_models.

predict(data, n_model)

Predict the class of the observation contained in data using the model number n_model.

confusion_matrix(real_class=None, pred_class=None, pred_prob=None, thresholds=array([0., 0., 0., 0., 0.]), norm=False, n_model=0)

Function for computing the confusion matrix.

Parameters:

• n_model (int) – Number of the machine learning methods that will be used for the classification.

• real_class (np.array) – Numpy array containing the real classes. If nothing is pass, the function will use the testset classes.

• pred_class (np.array) – Numpy array containing the predicted classes. If nothing is pass, the function will use the classes predicted for the testset.

• pred_prob (np.array) – Numpy array containing the predicted probabilities. If nothing is pass, the function will use the pred_class argument.

• thresholds (np.array) – Numpy array containing the thresholds that will be use for classifing the galaxies. Default = np.array([0.0, 0.0, 0.0, 0.0, 0.0])

• norm (Bool.) – Boolean indicating if the return confusion matrix should be normalized or not. Default = False

predict_class(data, n_model, verbose=1)

Function for predicting the orbital class.

Parameters:

• data (np.array) – Numpy array containing the features that will be used for the classification.

• n_model (int) – Number of the machine learning methods that will be used for the classification.

Return type:

array

Returns:

• Orbital class.

predict_prob(data, n_model)

Function for predicting the orbital class probability.

Parameters:

• data (np.array) – Numpy array containing the features that will be used for the classification.

• n_model (int) – Number of the machine learning methods that will be used for the classification.

Return type:

array

Returns:

• Probability for each galaxy of belonging to each orbital class.

split(split_seed)

Function for splitting the dataset into train and test set. It will run when instantiating a RogerModel object.

Returns:

• Tuple with the indices of the training and testing sets.

train(path_to_saved_model=None, path_to_save=None)

Function for training the machine learning methods.

pyROGER.models

pyROGER.models.list_saved_models()

Function for listing the available models.

Datasets