mlplot 0.0.1

No project description provided

mlplot

Machine learning evaluation plots using matplotlib and sklearn.

Install

pip install mlplot

ML Plot runs with python 3.5 and above! (using format strings and type annotations)

Contributing

Create a PR!

Plots

Work was inspired by sklearn model evaluation.

Classification

ROC with AUC number

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.roc_curve()

Calibration

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.calibration()

Precision-Recall

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.precision_recall(x_axis='recall')
eval.precision_recall(x_axis='thresold')

Distribution

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.distribution()

Confusion Matrix

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.confusion_matrix(threshold=0.5)

Classification Report

from mlplot.model_evaluation import ClassifierEvaluation
eval = ClassifierEvaluation(y_true, y_pred, class_names, model_name)
eval.report_table()

Regression

• Full report
  • Mean sqr error
  • Mean abs error
  • Target mean, std
  • R2
• Residual plot
• Scatter plot
• Histogram of regressor

Library Cleanup

• Try in a notebook
• Logging
• Default matplotlib setup
• Multi-model comparison
• Report to generate multiple plots at once

Development

Publish to pypi

python setup.py sdist bdist_wheel
twine upload --repository-url https://upload.pypi.org/legacy/ dist/*

Design

Basic interface thoughts

from mlplot.model_evaluation import ClassifierEvaluation
from mlplot.model_evaluation import RegressorEvaluation
from mlplot.model_evaluation import MultiClassifierEvaluation
from mlplot.model_evaluation import MultiRegressorEvaluation
from mlplot.model_evaluation import ModelComparison
from mlplot.feature_evaluation import *

eval = ClassifierEvaluation(y_true, y_pred)
ax = eval.roc_curve()
auc = eval.auc_score()
f1_score = eval.f1_score()
ax = eval.confusion_matrix(threshold=0.7)

• ModelEvaluation base class
• ClassifierEvaluation class
  • take in y_true, y_pred, class names, model_name
• RegressorEvaluation class
• MultiClassifierEvaluation class
• ModelComparison
  • takes in two evaluations of the same type