Skip to main content

skflex provides a suite of flexible utility functions for use with the sklearn library

Project description

skflex

FLEXIBLE FUNCTIONS ----- FAST PROCESSING AND EVALUATION

skflex provides a suite of utility functions for use with the sklearn library. The module primarily focuses on producing typical plots and metrics for evaluating machine learning models. It has been designed with flexability and customisation in mind to speed up workflows, and enhance comparative evaluation.

Installation and Import

pip install skflex

import skflex.skflex as skf

Functions

Functions currently included along with descriptions and default parameter settings.

refer to the Github repository for example images of plots.

roc_auc_plot

Accepts fitted model(s) and test data. It will then:

  • Calculate ROC
  • Calculate AUC
  • Plot ROC curve with AUC provided in the legend

Parameters:

  • models - fitted model objects. NOTE: Only models with a 'predict_proba' or 'decision_function' method are supported.
  • X_test - test feature set
  • y_test - test labels set
  • title - title for ROC curve plot
  • width - plot width
  • height - plot height
  • legend_size - size of plot legend

Default:

models, X_test = None, y_test = None, width = 14, height = 12, legend_size = 14, title='ROC Curve'

Example:

from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import LogisticRegression

model_1 = GaussianNB()
model_2 = LogisticRegression()

model_1.fit(X_train, y_train)
model_2.fit(X_train, y_train)

skf.roc_auc_plot(model_1, model_2, X_test = X_test, y_test = y_test, 
                title = 'Example ROC plot')

classifier_train_report

Accepts classifier models, training data, and test data. It will then:

  • Fit the model(s) to training data
  • Make predictions using test data
  • Produce classification report for comparison
  • Produce confusion matrix for comparison
  • Provide an ordered summary (ranked best to worst score) using given evaluation metric

Parameters:

  • models - model objects to be trained and evaluated
  • X_train - training feature set
  • y_train - training label set
  • X_test - test feature set
  • y_test - test label set
  • scoring - summary evaluation metric. Common classifier evaluation metrics including accuracy, f1, precision, and recall are supported. Refer to sklean scoring documentation for more information. Scoring methodologies should be passed as strings, for example, precision would be passed as scoring = 'precision'
  • title - title for output report

Default:

models, X_train = None, y_train = None, X_test = None, y_test = None, scoring = 'accuracy', title = 'Reports'

Example:

from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import LogisticRegression

model_1 = GaussianNB()
model_2 = LogisticRegression()

skf.classifier_train_report(model_1, model_2, X_train = X_train, y_train = y_train, 
                            X_test = X_test, 
                            y_test = y_test, 
                            scoring = 'accuracy', 
                            title = 'Example Report')

validation_plot

Accepts a model, a related hyper-parameter, a list of hyper-parameter values, training and test data, number of cross-validation folds, scoring methodology, as well as a plot title. It will produce a plot of the validation curve for the training and test data using the mean scores and standard deviations obtained through the cross-validation process.

Parameters:

  • model - model object
  • param - hyperparameter to be used to produce the validation curve
  • param_grid - hyperparameter values to be tested
  • X_train - training feature set
  • y_train - training label set
  • cv - number of cross-validation folds
  • scoring - scoring methodology used during cross-validation process
  • title - title for validation plot
  • width - plot width
  • height - plot height

Default:

model = None, param = None, param_grid = None, X_train = None, y_train = None, cv = 5, scoring = 'accuracy', width = 9, height = 9, title = 'Validation Curve'

Example:

from sklearn.naive_bayes import DecisionTreeClassifier

model_1 = DecisionTreeClassifier()
params = [5, 10, 15, 20, 30, 40, 50]

skf.validation_plot(model = model_1, param = 'max_depth', param_grid = params, 
                    X_train = X_train, 
                    y_train = y_train, 
                    title = 'Example Validation Curve')

train_val_test

Accepts a Pandas dataframe and will return a training, validation, and test set. Operates in a similar fashion to the sklearn train_test_split function by defining a percentage split for the training and validation sets (example 0.6 = 60%). The remainder is allocated to the test set.

Parameters:

  • data - dataframe to be split into a training, validation, and test set
  • class_labels - column in the dataframe containing class labels
  • train - percentage of data to be allocated to the training set
  • val - percentage of data to be allocated to the validation set
  • shuffle - if true, will shuffle the rows in the dataframe before splitting
  • random_state - if shuffle is ture, will set a random seed so that ordering produced by shuffle can be reproduced

Default:

data = None, class_labels = None, train = 0.6, val = 0.2, shuffle = True, random_state = None

Returns: X_train, y_train, X_val, y_val, X_test, y_test

Example:

X_train, y_train, X_val, y_val, X_test, y_test = skf.train_val_test(data = my_data, 
                                                                    class_labels = 'labels', 
                                                                    train = 0.6, 
                                                                    val = 0.2)

pca_scree_plot

Accepts data (array/dataframe), and number of principal components to be analysed. It will produce a scree plot of the cumulative variance explained.

Parameters:

  • data - dataset to be analysed
  • n_components - number of principal components to be analysed
  • width - width of plot
  • height - height of plot
  • legend_size - size of plot legend
  • scale_data - normalises data before analysis and plotting. If the data being passed has not yet been normalised, this parameter should be set as scale_data = True
  • title - plot title

Default:

data = None, n_components = None, width = 16, height = 10, legend_size = 12, scale_data = False, title = 'PCA Scree Plot'

Example:

from sklearn.preprocessing import scale

n_data = scale(my_data)

skf.pca_scree_plot(data = n_data, n_components = 70, title = 'Example PCA Scree Plot')

Requirements

  • Sklearn >= 0.24.1
  • Matplotlib >= 3.3.4
  • Pandas >= 1.2.4
  • Numpy >= 1.20.1

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

skflex-1.0.0.tar.gz (5.5 kB view hashes)

Uploaded Source

Built Distribution

skflex-1.0.0-py3-none-any.whl (5.9 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page