DeepRetail

Python package on deep learning AI and machine learning for Retail

This package is developed by the AI team at VIVES University of Applied Sciences and is used in our research on demand forecasting.

---

Getting started

Installation

1. Install python3.7+

2. Create a virtual env where you want to install:

   $> python3 -m venv retailanalytics
   

3. Activate the environment

   $> source retailanalytics/bin/activate
   

4. Install the package with pip

   $> pip install DeepRetail
   

Use hierarchical modelling

import pandas as pd
from DeepRetail.transformations.formats import transaction_df
from DeepRetail.forecasting.statistical import StatisticalForecaster

# Load
df = pd.read_csv('daily_data.csv', index_col=0)

# Get a sample 
sampled_df = df.sample(20)

# Convert to transaction
t_df = transaction_df(sampled_df)

# Define the parameters
freq = 'M'
h = 4
holdout = True
cv = 2
models = ['ETS', 'Naive']

# Convert columns to datetime
sampled_df.columns = pd.to_datetime(sampled_df.columns)

# Resample columns to montly frequency
sampled_df = sampled_df.resample('M', axis=1).sum()

# Define the forecaster
forecaster = StatisticalForecaster(models = models, freq = freq)

# Fit the forecaster
forecaster.fit(sampled_df, format = 'pivoted')

# Predict
forecast_df = forecaster.predict(h = h, cv = cv, holdout = holdout)
forecast_df.head()

Contributing

Contribution is welcomed!

Start by reviewing the contribution guidelines. After that, take a look at a good first issue.

Disclaimer

DeepRetail is an open-source package. We do our best to make this package robust and stable, but we do not take liability for any errors or instability.

Support

The AI team at VIVES University of Applied Sciences builds and maintains DeepRetail to make it simple and accessible. We are using this software in our research on demand forecasting. A special thanks to Ruben Vanhecke and Filotas Theodosiou for their contribution. The maintenance workflow can be found here.