pycewise 0.1.0

Piecewise linear regressions, based on model trees.

Pycewise

Installation

From a wheel (recommended)

pip install pycewise

Optional requirements

The main functionnality of this package (computing a segmented linear regression) can be used without any third-party code.

For additional features, the following packages should be installed (pip install <package_name>):

• numpy
• statsmodels
• jupyter
• matplotlib
• graphviz
• coverage
• mock
• palettable

Usage

Basic example:

from pycewise import *
import pandas

df = pandas.read_csv('test_data/ringrong_loopback.csv').groupby('size').mean().reset_index()
reg = compute_regression(df['size'], df['duration'], mode='log')
print(reg)

For more advanced usage, see the notebooks.