FORECASTING AT LEAST ONE FEATURE TO BE FORECASTED

The invention is directed to a computer-implemented method for forecasting at least one feature to be forecasted; comprising the steps:a. Providing a train network (10), wherein the train network (10) represents a plurality of nodes (12) in a graph which are interconnected by respective edges (14); wherein each node (12) of the plurality of the nodes represents a train entity; wherein each edge (14) of the plurality of the edges represents a relationship between the train entities (S1); b. Providing an electrical grid (20), wherein the electrical grid (20) represents a plurality of nodes (22) in a graph which are interconnected by respective edges (24); wherein each node (22) of the plurality of the nodes represents an electrical entity; wherein each edge (24) of the plurality of the edges represents a relationship between the electrical entities (S2); c. Combining the train network (10) and the electrical grid (20) using a combined hybrid graph topology (30)(S3); d. Forecasting the respective at least one feature to be forecasted using a trained machine learning model based on the combined hybrid graph topology (30)(S4); and e. Providing the at least one forecasted feature (S5). Further, the invention relates to computing unit and a computer program product.