Forecasting Sales and Developing a Newsvendor Model for Weather-sensitive Retail Products

<p>The sales of many retail products are affected by the weather. As the weather becomes more erratic due to climate change, retailers must respond by using weather information to help them plan their inventory. This thesis develops a sales forecasting model and an inventory planning model using weather information. The sales forecasting model is developed using several machine learning algorithms and several different sets of timeshifted weather features. Meanwhile, the inventory planning model is developed based on the Newsvendor model considering the demand-temperature relationship and solving the optimal order quantity. Both models are analyzed using sales data obtained from a large Canadian retailer. Using this data, important weather features are identified through analysis of the sales forecasting models. The results are then used in a numerical example and sensitivity analysis to demonstrate the proposed inventory planning model.</p>