| CPC G06Q 10/0637 (2013.01) [G06Q 10/06393 (2013.01)] | 12 Claims |
|
1. A computer implemented method comprising:
identifying, by a number of processor units, environmental data and organization parameters for organizations;
training, by the number of processor units, a machine learning model to predict emissions for the organizations using the environmental data and the organization parameters;
predicting, by the machine learning model, the emissions for the organizations;
identifying, by the number of processor units and an environmental analyzer, a group in groups of organizations for an organization of interest using organization parameters for the organization of interest;
determining, by the number of processor units and the environmental analyzer, environmental performance for the organizations in the group using the emissions predicted for the organizations in the group;
training, by the number of processor units, another machine learning model using previous recommendations of previous environmental projects as training data;
identifying, by the another machine learning model, a set of environmental projects for the organization of interest based on the environmental performance determined for the organizations in the group;
determining, by the number of processor units and the environmental analyzer, sector similarity loss between sectors for the organizations using a dependency graph of sectors based on activities of the organizations, wherein edges between nodes in the dependency graph of sectors have edge weights that indicate dependencies between the nodes connected by the edges;
determining, by the number of processor units and the environmental analyzer, organization similarity loss between the organization using organization parameters for the organizations;
clustering of organizations, by the number of processor units and the environmental analyzer, into the groups of organizations using a clustering machine learning model, the sector similarity loss and the organization similarity loss;
recommending, by the number of processor units and the environmental analyzer, in response to clustering the organizations into the groups of organizations, environmental projects; and
purchasing, by the number of processor units, in response to recommending environmental projects, a selected amount of renewable energy, wherein identifying, by the number of processor units, the set of environmental projects for the organization of interest comprises:
identifying, by the number of processor units, a set of candidate environmental projects for the organization of interest based on the environmental performance determined for the organizations in the group; and
refining, by the number of processor units, the set of candidate environmental projects for the organization of interest to reduce a negative impact on the environmental performance within a supply chain in which the organization of interest is located to form to form the set of environmental projects, wherein refining, by the number of processor units, the set of candidate environmental projects comprises:
modeling, by the machine learning model, upstream emissions and downstream emissions in the supply chain caused by the set of candidate environmental projects;
defining, by the number of processor units, an objective function for the set of candidate environmental projects taking into account the upstream emissions and the downstream emissions in the supply chain using the set of candidate environmental projects for the organization of interest;
performing, by the number of processor units, a counterfactual query using the objective function; and
selecting, by the another machine learning model, the set of environmental projects from the set of candidate environmental projects based on results of the counterfactual query.
|