ANALYZING UNCONTROLLABLE FACTORS THAT CAUSE DEFECTIVE PRODUCTS BY POISSON AND NEGATIVE BINOMIAL INAR(1) FOR FITTING MODEL

• Published in: Proceedings on engineering sciences (Online) Vol. 7; no. 1; pp. 1 - 10
• Main Authors: Kartika Gandhi, Herry, Marton, Ispany
• Format: Journal Article
• Language: English
• Published: University of Kragujevac 10.03.2025
• Subjects: forecasting; negative binomial inar1; noise factors; poisson inar1; product defect
• ISSN: 2620-2832; 2683-4111
• DOI: 10.24874/PES07.01.001

This study uses non-negative integer-valued first-order autoregressive time series models (INAR(1)), namely Poisson (PINAR) and negative binomial (NBINAR) ones, to analyze and to forecast the emergence of noise factors in the manufacturing production process that results in abnormal products. The use of the appropriate model is determined by using overdispersion analysis of the Poisson value and Index of Dispersion (ID). This research is based on data of break occurrences in 3 paper machines. The best model for machine 1 is NBINAR(1), while for machine 2 and 3 is PINAR(1). The results of goodness-of-fit (AIC and BIC) show that the error values of PINAR(1) and NBINAR(1) are lower than the Poisson and negative binomial iid cases by 2% (1377/1388 and 1401/1409). Proposed method is able to help production manager to identify the influence of uncontrollable factors (noise factors) that occur in the manufacturing area.