Machine learning for demand forecasting

AI Quick Summary

Operational forecasting allows supporting inventory decisions and set safety amounts.We provide state-of-the-art forecasting algorithms for demand data.We are authors of international publications on this topic. Our algorithm provides forecasts which are accurate and they scale efficiently over thousands of articles.They deal with smooth demand, intermittent demand and hierarchies of items.Starting from this expertise, we can develop a customized solution meeting the needs of an implementation partner. Operational forecasting supports inventory decisions and set safety amounts.We provide state-of-the-art forecasting algorithms for demand data of different types: smooth, intermittent and hierarchies.We are authors of international publications on this topic. Our algorithm provides forecasts which are accurate and they scale efficiently over thousands of articles.Starting from this expertise, we can develop a customized solution meeting the needs of an implementation partner.Most products are sold in units. When the sales have high volume, we refer to them as smooth demand.In recent years it has become common to store data of finer and finer granularities. This brings new challenges.A sufficiently fine granularity along location, product or time turns smooth demand data into intermittent demand, i.e., data that exhibit many zeros. This requires specialized forecasting methods.Often time series are obtained by aggregation. For instance, we might aggregate the sales of different types and brands of shoes in order to get the total sales of shoes; or we could aggregate the regional sales in order to get the national sales, etc. Such data are referred to as a hierarchical time series. The forecasts of hierarchical time series should be coherent; e.g., the forecasted sales for the different types and brand of shoes should match, when summed, the predicted total sales of shoes.We also have efficient algorithms for producing forecasts for hierarchies containing thousands of time series.Coherent forecasts allow taking aligned decisions at the different levels of the company, minimizing stockouts.

Project Idea Metadata

• Project Idea Name: Machine learning for demand forecasting
• Date: 2/7/2025 3:27:13 PM
• Administrators:
  • giorgio corani

Project Idea Description

Operational forecasting supports inventory decisions and set safety amounts.

We provide state-of-the-art forecasting algorithms for demand data of different types: smooth, intermittent and hierarchies.

We are authors of international publications on this topic.

Our algorithm provides forecasts which are accurate and they scale efficiently over thousands of articles.

Starting from this expertise, we can develop a customized solution meeting the needs of an implementation partner.

Most products are sold in units. When the sales have high volume, we refer to them as smooth demand.

In recent years it has become common to store data of finer and finer granularities.

This brings new challenges.

A sufficiently fine granularity along location, product or time turns smooth demand data into intermittent demand, i.e., data that exhibit many zeros. This requires specialized forecasting methods.

Often time series are obtained by aggregation. For instance, we might aggregate the sales of different types and brands of shoes in order to get the total sales of shoes; or we could aggregate the regional sales in order to get the national sales, etc. Such data are referred to as a hierarchical time series.

The forecasts of hierarchical time series should be coherent; e.g., the forecasted sales for the different types and brand of shoes should match, when summed, the predicted total sales of shoes.

We also have efficient algorithms for producing forecasts for hierarchies containing thousands of time series.

Coherent forecasts allow taking aligned decisions at the different levels of the company, minimizing stockouts.

Operational forecasting allows supporting inventory decisions and set safety amounts.

We provide state-of-the-art forecasting algorithms for demand data.

We are authors of international publications on this topic.

Our algorithm provides forecasts which are accurate and they scale efficiently over thousands of articles.

They deal with smooth demand, intermittent demand and hierarchies of items.

Starting from this expertise, we can develop a customized solution meeting the needs of an implementation partner.