Semester

Fall

Date of Graduation

2006

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Industrial and Managements Systems Engineering

Committee Chair

Majid Jaraiedi.

Abstract

An important aspect of forecasting system is monitoring the process for forecast accuracy. Tracking signals test is an effective detection method for occurrence of nonrandom changes. In this research, the relative efficiency of Cumulative Sum (CUSUM), Smoothed Error, Backward CUSUM, Autocorrelation and Parabolic CUSUM tracking signals were compared according to ARL1 criterion and percentage of trips detected within N periods. For simulating response to a step, 1,000 series of demands were generated, each covering run-ins of 40 periods and 75 periods after introduction of the step. The response of tracking signals to step sizes of 1sigma, 2sigma, 3sigma and a random step between 1sigma and 3sigma were studied at unbiased ARL of 25, 50 and 100. Smoothing parameters were varied from 0.05 to 0.2 in steps of 0.05. When the smoothing parameter is set equal to 0.05, 0.15 or 0.2 and other smoothing constants are 0.05, autocorrelation signal was the best choice. When the smoothing parameter is equal to 0.1 and other smoothing constants are 0.05, BCUSUM was the best choice when unbiased ARL=100; autocorrelation signal was the best choice when unbiased ARL=25 or 50. Autocorrelation signal is the best choice based on percentage of trips detected within N periods.

Share

COinS