System of dynamic optimization pricing by machine learning

Based on real retail data collected over eight months, a number of models are developed to determine the most efficient machine learning algorithm for dynamic pricing. The goal of the study is to determine the best algorithm that provides the minimum error and maximum accuracy of price forecasting for different transaction volumes. The main tasks included comparing the performance of several models: KNeighbors Regressor, Random Forest Regressor, Decision Tree Regressor, Linear Regressor, and Support Vector Regressor. It was found that the KNeighbors Regressor model showed the best results with a small number of transactions (MSE 0.0091, R2 0.72), while ensemble methods (Random Forest, Decision Tree) are best suited for large amounts of data. Instead, linear methods had significant deviations. The conducted research confirms the importance of artificial intelligence algorithms in the dynamic pricing strategy, demonstrating their ability to quickly process large amounts of data and take into account numerous factors of market demand.