brendanartley / regressio

A python library for univariate regression, interpolation, and smoothing.

Regressio is a python module for univariate regression, interpolation, and smoothing.

The available models are:

• Linear regression
• Ridge regression
• Linear spline
• Isotonic regression
• Bin regression
• Cubic spline
• Natural cubic spline
• Exponential moving average
• Kernel functions (Gaussian, KNN, Weighted average)

There are also functions implemented to generate data samples.

The available data generators are:

• Random walk
• Isotonic sample

Installation

Regressio is supported in Python 3.8+ and requires only NumPy and Matplotlib.

pip install regressio --upgrade

or

pip install git+https://github.com/brendanartley/Regressio

Example Usage

Cubic spline.

# Import modules + classes
from regressio.models import cubic_spline
from regressio.datagen import generate_random_walk
import numpy as np
import matplotlib.pyplot as plt

# Set figsize and seed
plt.rcParams['figure.figsize'] = (10, 5)
np.random.seed(0)

# Generate data sample
x, y = generate_random_walk(150)

# Fit model and plot result
model = cubic_spline(pieces=15)
model.fit(x, y, plot=True, confidence_interval=0.99)

Linear regression.

# Import modules + classes
from regressio.models import linear_regression
from regressio.datagen import generate_random_walk
import numpy as np
import matplotlib.pyplot as plt

# Set figsize and seed
plt.rcParams['figure.figsize'] = (10, 5)
np.random.seed(1)

# Generate data sample
x, y = generate_random_walk(100)

# Fit model and plot result
model = linear_regression(degree=5)
model.fit(x, y, plot=True, confidence_interval=0.95)

Exponential moving average.

# Import modules + classes
from regressio.models import exp_moving_average
from regressio.datagen import generate_isotonic_sample
import numpy as np
import matplotlib.pyplot as plt

# Set figsize and seed
plt.rcParams['figure.figsize'] = (10, 5)
np.random.seed(6)

# Generate data sample
x, y = generate_isotonic_sample(100)

# Fit model and plot result
model = exp_moving_average(alpha=0.2)
model.fit(x, y, plot=True, confidence_interval=0.90)

For more examples, navigate to the examples.ipynb file in this repository.

Contributions

We welcome all to contribute their expertise to the Regressio library. If you are new to open source contributions, this guide gives some great tips on how to get started.

If you have a complex feature in mind or find a large bug in the code, please create a detailed issue and we will get to work on it.

References

• Hyndman, R.J., & Athanasopoulos, G. (2021) Forecasting: principles and practice, 3rd edition, OTexts: Melbourne, Australia. OTexts.com/fpp3. Accessed July 2022.

• Kong, Qingkai, et al. Python Programming and Numerical Methods: A Guide for Engineers and Scientists. Academic Press, an Imprint of Elsevier, pythonnumericalmethods.berkeley.edu, Accessed July 2022.

• Li, Bao, (2022). Stat 508: Applied Data Mining, Statistical Learning: Stat Online. PennState: Statistics Online Courses, online.stat.psu.edu/stat508, Accessed July 2022.

• Brett, M. (2014, October 26). An introduction to smoothing. Tutorials on imaging, computing and mathematics. matthew-brett.github.io/teaching, Accessed July 2022.

BibText

@misc{Regressio,
      title={Regressio: A python module for univariate regression, interpolation, and smoothing}, 
      author={Brendan Artley},
      year={2022},
      publisher={GitHub},
      journal={GitHub repository},
      Howpublished = {\url{https://github.com/brendanartley/Regressio}}
}

regressio's Issues

Improve z-score calculation

Currently the only confidence intervals supported are those encoded in a dictionary. Need to implement a method that accepts any confidence interval between 0-1.

Add Kernel Average Smoother

Add License to Repo

Add prediction function for EMA and Gaussian Kernel

Split Interpolation and Regression

• First need to implement more interpolation methods (Cubic Spline, Natural Cubic Spline)
• Need to rename Isotonic regression to Isotonic Interpolation
• Separate interpolation functions from regression models (ex. regression.py, interpolation.py)

Add Gaussian Kernel Smoother

Add Lasso Regression

Add Ridge Regression

Fix examples.ipynb file

Exponential smoothing model uses the old notation. Need to update this file to the new class name

Add anomaly detection functionality

Add confidence interval to plots

Consider data storage in class

Consider storing the x, y, and smoothed values in the model for prediction interval calculation. This could result in easier confidence interval/prediction interval calculations.

Implement Cubic + Natural Cubic Spline

Fix cubic splines where splices == n-1

The current functionality takes the mean between two points when there is no point in the spline location. Need to fix this when this not the case.

Github Actions .YML Script

Add github actions that runs pytest script on push and pull requests.

Add Docstrings to Classes

Add more comprehensive docstrings to each class.

Add Nearest Neighbour Smoother