Forecasting demand reliably is integral for retailers for better inventory turnover leading to improved product availability and higher sales.
By using feature engineering on available data of past orders and inventory and then training model using ML to maximize profit and reduce prediction errors, it is possible to achieve reliable and accurate forecasts.
The python notebook demonstrates the above using data of 180 days with a 15-day ahead forecasting model.
A plotly dash application was built for visualizing the data and model results.
The application has been deployed on Heroku: https://demand-forecast-app.herokuapp.com/