Cross-sell řešení s využitím strojového učení v obchodní společnosti

The goal of many businesses is to increase sales and satisfy customer needs. One of the ways to increase these indicators in retail companies is to implement a recommendation system based on the cross-sell principle. The objectives of this paper are to design and implement a cross-sell solution using machine learning for a retail company. To achieve this objective, an analysis of the data status, needs and requirements of the trading company has been carried out. Also, the communication during the course of the project and its final status was arranged. Then a machine learning model based on neural networks with multiple inputs in the form of vectorized product data was prepared. Available historical orders data was used to train the model. Subsequently, the model was implemented in an API which was then deployed in a cloud environment to run independently of the corporate infrastructure. The CRISP-ML(Q) methodology was used to achieve the objectives of the work. The effectiveness of the finished solution was validated through testing of the cross-sell offers using the A/B/n method. The newly developed solution guaranteed the increase in the value of the average order compared to the static offer of random products when displaying the customer cart.