A Model for Forecasting the Computer Requirements for the Air Force Institute of Technology School of Systems and Logistics.
Abstract
This thesis analyzes and forecasts the usage pattern of the Air Force Logistics Command's CREATE computer system by the Air Force Institute of Technology School of Systems and Logistics. The analysis uses exponential smoothing in time-series forecasting, autocorrelation, bivariate correlation, and linear regression techniques to determine relationships among monthly aggregated computer usage parameters. The monthly aggregated computer parameters included batch jobs, CPU hours batch, CPU hours time-sharing, core hours, log-on hours, log-ons and lines CARDIN. While time-series forecasting can be satisfactorily used in most cases for forecasting purposes, autocorrelation analysis provided insight into the cyclic relationships between the data and time. The authors concluded that an unnecessarily long time-series can introduce inaccuracies into a forecast because the process of use of the system changes over time. Bivariate correlation analysis and linear regression demonstrated both simple and multiple linear relationships between the variables used. The authors concluded that these linear relationships could be used in defining the requirements for a new computing system to replace an existing system or part of a system. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1977
- Accession Number
- ADA047277
Entities
People
- David C. Johnston
- Herman H. Detjen
Organizations
- Air Force Institute of Technology