Overview
magi is a high level python wrapper around other time series forecasting libraries to allow easily parallelized univariate time series forecasting in python by using dask delayed wrapper functions under the hood. In particular, the library currently supports wrappers to R forecast library and facebook’s prophet package
Usage
This is how easy it is to clean, forecast, and then plot accuracy metrics for 100 time seres using the auto arima model from R forecast package
Importing libraries, generate dataframe of series for example, and start local dask cluster
from magi.core import forecast
from magi.plotting import fc_plot, acc_plot
from magi.utils import gen_ts
from magi.accuracy import accuracy
from dask.distributed import Client, LocalCluster
import dask
cluster = LocalCluster()
client = Client(cluster)
df = gen_ts(ncols=100)cleaning and forecasting for 100 series in parallel, then calculate and plot accuracy metrics by series
fc_obj = forecast(time_series=df,forecast_periods=18,frequency=12)
forecast_df = fc_obj.tsclean().R(model='auto.arima(rdata,D=1,stationary=TRUE)',fit=True)
acc_df = accuracy(df,forecast_df,separate_series=True)
acc_plot(acc_df)Use Cases
What this package should be used for
forecasting for 1 or more Univariate Time Series
forecasting using many different time series models in parallel with minimal effort
wrapper for R forecast library to implement those models in python workflow
wrapper around Prophet library to provide easier data framework to work with
single source of access for many different time series forecasting models
What this package should NOT be used for
Multivariate Time Series data. If you have multiple x variables that are correlated with your response variable, I’d suggest simply using regression with lags and seasonal variable to account for autocorrelation in your error
Data exploration - The time series analysis step is much more suited to using the R forecast package directly
Dependencies
dask
distributed
plotly
cufflinks
rpy2 (& forecast package >=8.3 installed in R)
fbprophet
Installation
$ pip install magiDocumentation
Documentation is hosted on Read the Docs.
Disclaimer
This package is still very early in development and should not be relied upon in production. Everything is still subject to change
1 maintainer
Release files for magi 0.0.17
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| magi-0.0.17.tar.gz | 11.2 kB | Details |
Release files / magi-0.0.17.tar.gz
| Download URL | magi-0.0.17.tar.gz |
|---|---|
| Size | 11.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
87ebb3f9914d69896d882a8a2d1d30dd44e695fa5d4b887d22466c880e66b3bd
|
|
BLAKE2b-256 checksum How to use checksums |
bf76422823fdbe47e4b16b9d3ed014a0c4bef12d324896008861520971b1a7ae
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
