Abstract
This study aims to investigate the contributions of promotional marketing activities, historical demand and other factors to predict, and develop a big data-driven fuzzy classifier-based framework, also called "demand-driven forecasting," that can shape, sense and respond to real customer demands. The availability of timely information about future customer needs is a key success factor for any business. For profit maximization, manufacturers want to sense demand signals and shape future demands using price, sales, promotion and others economic factors so that they can fulfil customer's orders immediately. However, most demand forecasting systems offer limited insight to manufacturers as they fail to capture contemporary market trends, product seasonality and the impact of forecasting on the magnitude of the bullwhip effect. This paper aims to improve the accuracy of demand forecasts. In order to achieve this, a back-propagation neural network-based model is trained by fuzzy inputs and compared with benchmark forecasting methods on a time series data, by using historical demand and sales data in combination with advertising effectiveness, expenditure, promotions, and marketing events data. A statistical analysis is conducted, and the experiments show that the method used in the proposed framework outperforms in optimality, efficiency and other statistical metrics. Finally, some invaluable insights for managers are presented to improve the forecast accuracy of fuzzy neural networks, develop marketing plans for products and discuss their implications in several fields.