Automatic classification of categories for Great Hobbies Inc. based on Vendor Information

This paper focuses on developing a product code classification system for Great Hobbies, a retail company specializing in the sale of hobby products. The system was built to provide a quick and efficient way to categorize the large number of products that the company offers. The pairing of products with the category is done by humans taking into account structured and unstructured data. Examples of unstructured data are name and product descriptions; structured data consist of the vendor’s product code. There are over 200 different combinations of department and product codes in Great Hobbies, which makes this pairing a time-consuming process. A machine learning-based approach was used to build the classification models. Multiple models were developed and evaluated based on their accuracy and time complexity. The best model was selected and deployed as a component of a decision support software system that runs on the Shopware platform. The system was designed with three main components: a front-end application, a web API, and backend workers. Results from the implementation show that the models have achieved a high degree of accuracy in classifying product codes into their respective categories.