peshbeen 0.0.1

No project description provided

peshbeen

peshbeen is a easy-to-use Python library designed for time series forecasting. It provides a simple and intuitive interface for building, training, and evaluating forecasting models. It automatically handles data preprocessing specific to time series data, making it easier for users to focus on model development and evaluation. Peshbeen supports a variety of forecasting models, including traditional statistical models and modern machine learning approaches, including Linear Regression, Random Forest, XGBoost, LightGBM, CatBoost, and Cubist.

Why peshbeen

• Automated Detrending Engine: Seamlessly handles non-stationary data using global linear trends, piecewise linear regression (for structural breaks), or ETS (Exponential Smoothing) to capture evolving trend dynamics.

• Broad Model Support: Out-of-the-box compatibility with the leading tree-based and boosting architectures, including XGBoost, LightGBM, CatBoost, Random Forest, and AdaBoost.

• Intelligent Optimization: Leverages Bayesian Optimization (Hyperopt) to not only tune machine learning hyperparameters but also to optimize smoothing parameters for ETS detrending.

• Uncertainty Quantification: Goes beyond simple point forecasts with a suite of probabilistic approaches:

  • Empirical Distributions: Leveraged via boosting.

  • KDE (Kernel Density Estimation): For non-parametric forecast distribution modeling.

  • Correlated Error Modeling: Accounting for error dependencies across the forecast horizon.

  • Conformal Prediction: Distribution-free uncertainty estimates.

Install in Development mode

# make sure  package is installed in development mode
$ pip install -e .

# make changes under nbs/ directory
# ...

# compile to have changes apply to 
$ nbdev_prepare

Installation

Install latest from the GitHub [repository][repo]:

$ pip install git+https://github.com//.git

or from [conda][conda]

$ conda install -c  

or from [pypi][pypi]

$ pip install 

[repo]: [docs]: https://.github.io// [pypi]: https://pypi.org/project// [conda]: https://anaconda.org//

Documentation

Documentation can be found hosted on this GitHub [repository][repo]’s [pages][docs]. Additionally you can find package manager specific guidelines on [conda][conda] and [pypi][pypi] respectively.

[repo]: [docs]: https://.github.io// [pypi]: https://pypi.org/project// [conda]: https://anaconda.org//

How to use

Fill me in please! Don’t forget code examples:

1+1

2

Models

API details

Peshbeen is …

from peshbeen.metrics import RMSE, MAE

MAE([3, -0.5, 2, 7], [2.5, 0.0, 2, 8])

np.float64(0.5)

from peshbeen.datasets import load_nhs_attend
# split the data into train and test sets
train = load_nhs_attend[:-30]
test = load_nhs_attend[-30:]

from peshbeen.models import ml_forecaster

# import linear regression from sklearn
from sklearn.linear_model import LinearRegression
ml_linear = ml_forecaster(model=LinearRegression(),
              target_col='attend', lags = 7)

ml_linear.fit(train)

ml_linear.forecast(H=30)

array([430796.7041251 , 428410.43400966, 426821.4107408 , 425222.28238429,
       424326.81261509, 423662.41365795, 423223.09618412, 422821.17942991,
       422415.76664768, 422068.11922156, 421747.36540958, 421464.13994277,
       421222.63819927, 421020.25542666, 420849.47190517, 420702.43079703,
       420576.84362351, 420469.04045579, 420375.80944133, 420295.35317043,
       420226.20137003, 420166.84067688, 420115.81185478, 420071.98653527,
       420034.37835101, 420002.07038525, 419974.29777864, 419950.43050693,
       419929.92365245, 419912.30021165])

# plot the forecast against the actual values
import matplotlib.pyplot as plt
plt.figure(figsize=(10, 6))
plt.plot(train.index, train['attend'], label='Train')
plt.plot(test.index, test['attend'], label='Test')
plt.plot(test.index, ml_linear.forecast(H=30), label='Forecast')
plt.legend()
plt.show()