lib package

Submodules

lib.data_processing module

lib.data_processing.copy_upsample(X, y, over_sampling)

Apply upsampling on minority data

Parameters:

• X (pandas DataFrame) – Training Features
• y (pandas Series) – Training Features

Returns:

resampled data

Return type:

tuple

lib.data_processing.fill_missing_knn(df, na_column, n_neighbors=10, algorithm='ball_tree')

fill missing data using k Nearest Neighbors

Parameters:

• df (pandas DataFrame) – Data Frame
• na_column (list) – column with nans
• n_neighbors (int) – number of neighbors
• algorithm (string) – nearest neighbors algorithm

Returns:

dataframe without nans

Return type:

pandas DataFrame

lib.data_processing.k_fold_prediction(X, y, n_splits, model, reg_params, fit_parameters, upsample_kwargs)

performs train-predict k fold

Parameters:

• X (pandas DataFrame) – Training Features
• y (pandas Series) – Training Labels
• n_splits (int) – number of k splits
• model (object) – model with fit and predict methods
• reg_params (dictonary) – model kwargs
• fit_parameters (dictonary) – model fit kwargs
• upsample_kwargs (dictonary) – syntetic_sampling kwargs

Returns:

predictions

Return type:

pandas Series

lib.data_processing.ohe(df, columns, drop_first=True)

apply OneHotEncoder

Parameters:

• df (pandas DataFrame) – Data Frame
• columns (list) – column with nans
• drop_first (boolean) – drop first ohe columns

Returns:

dataframe with ohe

Return type:

pandas DataFrame

lib.data_processing.read_data(path, remove_nans=True, apply_ohe=True)

read and transform data

Parameters:

• path (string) – path to csv file
• remove_nans (boolean) – remove nans from DataFrame
• apply_ohe (boolean) – apply OneHotEncoder on categorical data

Returns:

clean data

Return type:

pandas DataFrame with transformations

lib.data_processing.syntetic_sampling(X, y, over_sampling, under_sampling)

Apply Synthetic Minority Oversampling Technique (SMOTE) to tn unbalanced class

Parameters:

• X (pandas DataFrame) – Training Features
• y (pandas Series) – Training Features

Returns:

resampled data

Return type:

tuple

lib.data_processing.train_test_sample(X, y, test_size, upsample_type=None, over_sampling=None, under_sampling=None)

Splits into train and test samples and applies transformations to the train sample

Parameters:

• X (pandas DataFrame) – Training Features
• y (pandas Series) – Training Features
• test_size (float) – test size from split, defaults to None
• upsample_type (float) – sampling method
• upsample_type – oversample rate
• under_sampling (float) – undersample rate

Returns:

train and test split

Return type:

tuple

lib.model module

class lib.model.BinaryClassifier

Bases: object

Binary Classifier Model

dump(path=None)

export model to file

Parameters:path (string) – path to dump model, defaults to None

fit(X, y)

fit binary classifier model

Parameters:

• X (pandas DataFrame) – Training features
• y (pandas Series) – Training labels

k_fold_prediction(X, y, n_splits=5)

performs train-predict k fold

Parameters:

• X (pandas DataFrame) – Training Features
• y (pandas Series) – Training labels
• n_splits (int) – number of k splits, defaults to 5

Returns:

k fold predictions

Return type:

pandas Series

load(path=None)

laod model from file

Parameters:path (string) – path to dump model, defaults to None

predict(X)

predict on data

Parameters:X (pandas DataFrame) – data to predict Returns:predictions Return type:pandas Series

class lib.model.MultiClassifier

Bases: object

Multiclass Classifier Model

dump(path=None)

Parameters:path (string) – path to dump model, defaults to None

fit(X, y)

fit multiclass classifier model

Parameters:

• X (pandas DataFrame) – Training features
• y (pandas Series) – Training labels

load(path=None)

laod model from file

Parameters:path (string) – path to dump model, defaults to None

predict(X)

predict on data

Parameters:X (pandas DataFrame) – data to predict Returns:predictions Return type:pandas Series

lib.utils module

lib.utils.Evaluate(true_label, predicted_label, predicted_prob, labels)

Plot confusion Matrix and displays accuracy f1 and roc_auc scores

Parameters:

• true_label (array) – ground truth values
• predicted_label (array) – predicted values
• predicted_prob (array) – probability for each predicted class
• labels (list) – list containing label strings

lib.utils.arg_nearest(array, value)

Find index of nearest value for a given number

Parameters:

• array (array) – numpy array
• value (float) – desired value

Returns:

index

Return type:

int

lib.utils.plot_confusion_matrix(cm, labels, suptitle='Confusion Matrix')

_subplot_cm wapper - Plot normalized and not normilized confusion matrix

Parameters:

• cm (array) – confusion matrix array
• labels (list) – list containing label strings
• suptitle (string) – plot title, defaults to Confusion Matrix

lib.utils.plot_precision_recall(y_true, preds_proba)

Plot precision recall curves

Parameters:

• y_true (array) – ground truth values
• preds_proba (array) – probability for positive predicted class