Clustering and e-Commerce: Towards a Crossroads in a Particular Context: Categorization of Amazon products problematic in intermediation agencies A new context for the use of clustering in e-commerce

In all marketplaces, including Amazon, knowing how to add your product in the most appropriate category is one of the determining factors for its sale. However, this work of categorization remains a rather long process, which requires a lot of research on the platform, without being guaranteed that the category finally chosen is really the best adapted to its product. This becomes even more complex in the case of an intermediation agency like Bizon, which manages hundreds of accounts on Amazon for various clients, where it must add a large quantity of products, making sure to choose the right categories. Moreover, in such a case, the process can require a lot of patience, as it involves multiple exchanges between the seller (the agency's customer), the integrator (the person who manages the customer's account in the agency) and the platform (Amazon), which can slow down the whole process considerably, and consequently, generate possible frustrations. So, this paper introduces our research work which focused on the optimization of this categorization process. More technically, we used thousands of data from our best-selling products to train a clustering model capable of suggesting/predicting the best category for a product based on keywords. This approach has actually enabled us to eliminate the passing of information between the three actors, in this case reducing the process from days to seconds. In the rest of this paper, after a presentation of this new context of using clustering in e-commerce, we make a detailed presentation of the problem and a state of the art. The end of the paper is devoted to the definition of our experimentation protocol and the presentation of the first results.