Multivariate Time Series Analysis for Optimum Production Forecast: A Case Study of 7up Soft Drink Company in Nigeria

Abstract

This study focuses on the establishment of an optimum forecast model that predicts future production trends of 7UP Bottling company. Sixty (60) months time series data of 7UP bottling company were used after ascertaining the presence of seasonal variation and trend components of the data to establish the multidimensional forecast model. Predictive Production rate model was developed using a general multivariate regression equation form. The monitoring schemes show values of MSE and MAD as 0.0177 and 0.0658 respectively giving a tracking signal of 0.0. These values established the multivariate forecast model as optimum approach in tracking demand and production trends in a production setup. The value of the standard deviation of distribution of errors of 0.0823 estimated with MAD also confirms the authenticity of this model. The responses shown in the graphics of this study clearly explains the mixed time series which definitely contains seasonal variation and trend components as established in this study. Also the coefficient of determination of 0.957956 explains about 97% fitness of the established model to production data. The trend component associated with Copyright © IAARR, 2010: www.afrrevjo.com 277 Indexed African Journals Online: www.ajol.info time variable (Mtncod) causes production to increase by 0.002579KG/Month. Finally, this work adds to the growing body of literature on data-driven production and inventory management by utilizing historical data in the development of useful forecasting mathematical model. Introduction A large assortment of forecasting techniques has been developed over the years past ,which has naturally led to studies comparing their forecasting abilities.The comparisms are often part of a search for the best extrapolation technique but compiled results were mixed and often contradictory (Narasimhan,1995). Again combinig forecasts from two or more techniques (such as simple averaging) can dramatically improve forecast accuracy (Amstrong, 1994, Bates, 1969, Newbold and Granger, 1974 and Whinkler and Makridakis, 1984). Vonderembse and White (1991) also recognized the factors influencing the time series to be associated with secular trend that reflects forces that are responsible for growth or decline over a long period of time, seasonal variation, that reflect forces that act periodically in a fixed period of one year or less, cyclical fluctuations, that occur periodically in a fixed period of more than one year and random fluctuations. This study looks at multiple linear regression model as a model to take care of many influencing factors in time-series trends since many independent variables and mixed factors are involved in the making of a product.