Predictive business operations management

  • Authors:
  • Malu Castellanos;Norman Salazar;Fabio Casati;Umesh Dayal;Ming-Chien Shan

  • Affiliations:
  • Hewlett-Packard, Palo Alto, CA;Hewlett-Packard, Palo Alto, CA;Hewlett-Packard, Palo Alto, CA;Hewlett-Packard, Palo Alto, CA;Hewlett-Packard, Palo Alto, CA

  • Venue:
  • DNIS'05 Proceedings of the 4th international conference on Databases in Networked Information Systems
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

Having visibility into the current state of business operations doesn't seem to suffice anymore. The current competitive market forces companies to capitalize on any opportunity to become as efficient as possible. The ability to forecast metrics and performance indicators is crucial to do effective business planning, the benefits of which are obvious – more efficient operations and cost savings, among others. But achieving these benefits using traditional forecasting and reporting tools and techniques is very difficult. It typically requires forecasting experts who manually derive time series from collected data, analyze the characteristics of such series and apply appropriate techniques to create forecasting models. However, in an environment like the one for business operations management where there are thousands of time series, manual analysis is impractical, if not impossible. Fortunately, in such an environment, extreme accuracy is not required; it is usually enough to know whether a given metric is predicted to exceed a certain threshold or not, is within some specified range or not, or belongs to which one of a small number of specified classes. This gives the opportunity to automate the forecasting process at the expense of some accuracy. In this paper, we present our approach to incorporating time series forecasting functionality into our business operations management platform and show the benefits of doing this.