Offer analysis and modelling in the automotive market

When analyzing the automotive industry, a complex aspect is the fact that vehicle versions enter and leave the market on the basis of the production by car manufacturers and the purchase by customers. Also, the composition of the offering in terms of equipment, power and other characteristics can change significantly from manufacturer to manufacturer, and in time within the same model. The aim of this thesis work is then to highlight areas of market offering constant over time, composed by vehicles similar to each other, in order to have information about which new comparable vehicles can be a substitute to the ones going off the market. The chosen technique is a combination of clustering of the initial data to group vehicles and classification to map the new vehicles to the existing clusters, with however also an additional clustering step updated at every time data change, considering special conditions like the introduction of never seen kind before vehicles that do not fit the model. K-Means and Agglomerative Hierarchical clustering where opportunely designed and tuned to this specific use case; concurrently a Large Language Model provided by Azure OpenAI Service was explored to evaluate performances and benefits. After tuning and evaluating the models, evaluation with company subject matter experts was conducted to define the effectiveness and applicability towards the objective.