Retail Demand Forecasting using Machine Learning Algorithms

Understanding how to forecast a product's sales and demand is crucial for businesses that sell goods. Knowing how much demand will be in a given time gives them many benefits and gains. Many methods have been developed and used for demand forecasting from past to present. If we divide the methods used into two, traditional and machine learning methods are used for demand forecasting. We can say that traditional methods have left their place to machine learning due to less and slow data processing. Machine learning methods have the ability to process a lot of data faster and analyze the data it uses and provide a more accurate prediction by identifying hidden patterns in the data. The problem here is that there is no one "onesize-fits-all" prediction algorithm. Typically, demand forecasting features consist of several machine learning approaches. Therefore, the choice of machine learning models depends on many factors such as business goal, data type, data quantity and quality, forecast time. Therefore, the main problem here is to determine which algorithm will be used with which parameters. In this study, different machine learning methods and parameters was used and compared to select the most suitable machine learning algorithm and parameters according to the selected data set and provide more accurate predictions. Algorithms such as time series, linear regression, random forest was studied and external factors such as seasonal, regional and economic factors was used as parameters. The algorithm with the best results will be chosen from models with or without external factors. Keywords: Machine Learning, Demand Forecasting, Regression, Time Series