Vytvoření doporučovacího systému na bázi neuronových sítí v oblasti e-commerce

Nowadays, more and more companies are engaging in online trading, leading to a gradual competition increase in this area. Due to the rising competition, companies are compelled to attract customers as much as possible to maximize their sales. One way they can achieve this is through cross-selling, which involves offering products related to the customer's purchase. To recommend the right products, recommendation systems are created. In the field of machine learning, there are many ways to create these systems. The most advanced and powerful forms of them use neural network technology. This thesis focuses on creating a neural network-based recommendation system to enhance online sales by providing product recommendations derived from the contents of the customer's current shopping basket. The main goal of the thesis is to create a recommendation system capable of predicting items in which the customer could show a potential interest, based on the current structure of the customer's basket, while exclusively recommending items that are not already present in the basket. To fulfill the main objective, it is necessary to meet these steps. •Conducting research in the field of modern recommendation systems and the mentioned neural networks •Creation of a recommendation system for Rotorama company products to support cross-selling •Testing and evaluating the success of the created system Ensuring knowledge about modern recommendation systems is addressed through literary research. To create a recommendation system, the Aurélien Géron control list is employed. The success of the created system is evaluated through the expert knowledge of Rotorama company employees. The thesis's contribution represents the exploration of a new machine learning approach to solve the recommendation problem in the field of online trading, resulting in a comprehensive guide for the development of recommendation systems. The output of the thesis is a functional recommendation system prepared for implementation, including the results of its testing.