Simülasyon ile elde edilmiş perakende datasının istatistik ve veri madenciliği yöntemleri ile talep tahmini

Primary research purpose of this thesis is to evaluate statistical and data mining techniques for demand forecasting in the presence of promotions. A consumer choice model is developed by modifying present models of consumer choice in marketing literature for grocery retail industry. Data generation task is carried out according to the developed consumer choice model. Data is generated in a retail environment which has a single category, multiple products, multiple product attributes and multiple customer segments. Marketing drivers, such as price discounts, advertisement and feature displays are in data generation. Forecasting is performed over generated data by using both traditional statistical techniques, such as exponential smoothing and regression and recent data mining techniques, such as support vector machine regression and regression tree. Consequently, forecasting results of several techniques are compared according to accuracy of the SKU demand forecasts, simplicity in terms of parameter estimation and forecasting performance in new SKU entered situations. Finally, preferred methods for different data conditions are explained.