Analyzing Collaborative Forecast and Response Networks

TL;DR

This paper introduces analytical methods using Gaussian Graphical Models and Continuum Canonical Correlation to analyze and visualize information flow in collaborative forecasting networks, revealing inefficiencies and aiding forecasting practices.

Contribution

It presents novel analytical tools for understanding information flow in collaborative forecasting networks, including visualization and simplification techniques.

Findings

Identified inefficiencies in information flow within forecasting networks

Developed visualization methods for network analysis

Implemented tools in forecasting software for practical use

Abstract

Collaborative forecasting involves exchanging information on how much of an item will be needed by a buyer and how much can be supplied by a seller or manufacturer in a supply chain. This exchange allows parties to plan their operations based on the needs and limitations of their supply chain partner. The success of this system critically depends on the healthy flow of information. This paper focuses on methods to easily analyze and visualize this process. To understand how the information travels on this network and how parties react to new information from their partners, this paper proposes a Gaussian Graphical Model based method, and finds certain inefficiencies in the system. To simplify and better understand the update structure, a Continuum Canonical Correlation based method is proposed. The analytical tools introduced in this article are implemented as a part of a forecasting solution software developed to aid the forecasting practice of a large company.