The AI model is composed of a neural network which contains thousands of parameters. It works by learning the pattern of pricing behavior on available combinations of attributes that define a carbon project (standard, project activity, geography, contribution to SDGs, vintage).
Once pricing patterns are learned as per the iterative training of the neural network, the model is capable of working out pricing on all possible combinations of those input attributes.
In the learning process, the AI is fed a dataset with transaction prices and market quotes from different dates. This dataset contains a diverse set of combinations in the input attributes, including projects that have certified SDGs, projects that only have self-reported SDGs, and SDG contributions assessed by VAI’s natural language processing engine.
For instance, the model learns the patterns in value of a Gold Standard renewable energy project in India with and without certified SDG “8”, among various other combinations of those attributes. In this training process, the model ends up capturing the full inherent probability distribution for those attributes combinations. This learning process happens simultaneously against the values of liquid instruments and other observable reference values, called factors or benchmarks.