Associative Knowledge Graph of the American Car Market
The case involves developing and validating a recommendation system and car price predictor using a Knowledge Graph, which replaces generic ML models and data structures, to represent the American used car market.
Learn more
Exploring the power of Multi-Associative Graph structures in automotive R&D
Our client is the research branch of a large automotive corporation. They are solely responsible for R&D activities for the automotive manufacturer.
Together with their researchers, we have decided to validate and test the concept of using a Multi-Associative Graph Structure, or simply a Knowledge Graph, as a replacement for generic ML models and data structures to represent the American car market – focusing on the used cars currently being offered and their associations with publicly available datasets regarding vehicles in general.
We have developed a recommendation system and car price predictor based on the graph.
For that project, the used car data from cars.com (over a million vehicles of different brands) was gathered to build a knowledge graph. The data was also associated with other available datasets, including information regarding CO2 emissions or charger availability and locations in the US.
Finally, we have validated the performance of knowledge graphs compared to regular ML tools and algorithms: regression, clustering, and pattern mining.
