Aggregate production plan based on a neural network forecasting model for premium grocery items in Bangladesh

Abstract

Demand fluctuation, shorter shelf life, and business competition make the grocery market one of the most vulnerable markets. In-order to overcome this situation, accurate demand forecasting is essential. Considering current situation, this study aims to develop a forecasting model considering the factors that influence the demand for premium grocery items in Bangladesh and also develop an optimized aggregate model considering the forecasted demand. The model is developed for two premium grocery items. In order to identify the factors, a structured questionnaire based survey has been conducted. From the survey, 12 factors are identified that influence demand. The most influencing factors are identified by using factor analysis. The factors are price, per capita income, and number of competitors, market share, product quality, product availability, seasonal variations, festival season, and promotional activities. Considering these factors, a feed- forward neural network forecasting model with the back-propagation algorithm is prepared. The selected factors are used as input for the model and demand data including the factors that have collected and used as input. Two years of data have used to train the model. Using the model, future 12 months demand is calculated. The model is developed for two premium grocery items. Based on the result, an optimized aggregate plan is developed.