KAYA, Yılmaz Siirt Üniversitesi - Mühendislik Mimarlık Fakültesi - Bilgisayar Mühendisliği Bölümü, Turkey , YEŞİLOVA, Abdullah Yüzüncü Yıl Üniversitesi - Ziraat Fakültesi - Zootekni Bölümü, Turkey

INVESTIGATION OF E-MAIL TRAFFIC BY USING ZERO-INFLATED REGRESSION MODELS

34083

Based on count data obtained with a value of zero may be greater than anticipated. These types of data sets should be used to analyze by regression methods taking into account zero values. Zero- Inflated Poisson (ZIP), Zero-Inflated negative binomial (ZINB), Poisson Hurdle (PH), negative binomial Hurdle (NBH) are more common approaches in modeling more zero value possessing dependent variables than expected. In the present study, the e-mail traffic of Yüzüncü Yıl University in 2009 spring semester was investigated. ZIP and ZINB, PH and NBH regression methods were applied on the data set because more zeros counting (78.9%) were found in data set than expected. ZINB and NBH regression considered zero dispersion and overdispersion were found to be more accurate results due to overdispersion and zero dispersion in sending e-mail. ZINB is determined to be best model according to Vuong statistics and information criteria.

51

E , mail traffic , Zero inflated models , Zero inflated data sets , Data mining.

Anadolu University Journal of Science and Technology. A : Applied Sciences and Engineering

63