In planning, operating and developing mobile data networks, one crucial input is therntelecommunication demand that includes number of subscribers and their requiredrnservice data rates. These numbers should be predicted accurately for optimal planningrnand to capture the needs of the subscriber thereby creating customer satisfaction.rnEthio-telecom, the sole telecom service provider in Ethiopia, has recently introducedrndifferent service charging systems that include flat rate and package-based datarnservices. This has increased the number of subscribers who are using these services,rnwhich in turn has led to a substantial increase in data traffic, and hence, a burden on thernexisting infrastructure. Such increases in demand should be considered in planningrnphases, where proper forecasting of the data demand growth is one integral input forrnthe planning. Based on the available information, the current data growth forecastrnpractice being employed by Ethio-telecom is mainly based on marketing information.rnThis thesis presents Seasonal Auto-Regression Integrated Moving Average (SARIMA)rnmodel as an alternative way of forecasting Universal Mobile TelecommunicationrnSystem (UMTS) mobile data traffic taking the city of Addis Ababa as a case study. Thernapproach in this thesis involves investigating the past UMTS data traffic load collectedrnfrom the core network to find an appropriate model which describes the inherentrnstructure of the UMTS data-traffic and forecast the future data traffic load. With thisrnforecasting model, it is observed that the expected monthly data traffic per user forrnsmart phones can reach up to 7GB as compared to the current 1GB cap. As the firstrnpractice (to the best of our knowledge) for data forecasting using available data inrnEthio-telecom, it is hoped that the approach shown here will be useful for subsequentrninfrastructure expansion planning in a way that guarantees better customer satisfaction.rnKey Words: ACF, forecasting, PACF, SARIMA, Seasonality, Trend, UMTS, UMTS Datatraffic.rnPage