رضوانه اقبالي

Sales forecasting in the retail industry is one of the important and influential factors in this sector. Given the competitive market in communities and the significance of customers in this industry, sales forecasting helps retailers understand product demand and meet customer needs. This thesis examines the prediction of retail product sales data using machine learning techniques and analyzes the impact of holidays on retail sales. To explore the topics of this research, a dataset from a large chain retail store in southern Tehran has been studied. This dataset was prepared in Persian for a five-month period in 2019. This research specifically focuses on forecasting sales for four products. Four methods were used for forecasting: Random Forest, Linear Regression, K-Nearest Neighbors (KNN), and ARIMA as time series methods. In the time series approach, two scenarios were considered: one with the holiday feature and one without it, showing that including the holiday feature improves forecasting results. Finally, by comparing the methods used for forecasting and eva‎luating their performance metrics, the best method for predicting sales data was identified. According to this research, the ARIMA algorithm was recognized as the top algorithm, achieving the lowest mean squared error with a 29% reduction in error compared to the second-ranked algorithm, which was Linear Regression.

پيشبيني فروش , خرده فروشي , فروشگاه زنجيره اي , يادگيري ماشين , تعطيلات

Sales Forecasting , Retail , Chain Stores , Machine Learning , Holiday

Rezvaneh Eqbali

Farnaz BarzinPour