Method for decision-making regarding a decision in an environment by means of a data processing system and a corresponding data processing system
A method for decision-making regarding a decision in an environment by a data processing system in view of multiple different objectives includes: collecting information within the environment, describing the information in at least one temporal knowledge graph (TKG), forecasting a future development of one future state or more future states at a future time or more future points in time, under different decisions by the at least one TKG. The method further describes each resulting future state/decision combination by a corresponding temporal knowledge graph, rates an adherence of each forecasted future state to each objective of the multiple different objectives, considers a trade-off between the objectives for decision-making in a time-aware manner; and provides the decision.
1 . A method for decision-making regarding a decision in an environment by a data processing system, wherein decision-making is performed in view of multiple different objectives, the method comprising:
collecting information within the environment;
describing the information in at least one temporal knowledge graph (TKG);
forecasting a future development of one future state or more future states at a future time or more future points in time, under different decisions by the at least one temporal knowledge graph;
describing each resulting future state/decision combination by a corresponding temporal knowledge graph;
rating an adherence of each forecasted future state to each objective of the multiple different objectives;
considering a trade-off between the different objectives for decision-making in a time-aware manner; and
providing the decision.
2 . The method according to claim 1 , wherein the information is collected by observation and interaction.
3 . The method according to claim 1 , wherein the forecasting step relies on a machine learning (ML) model.
4 . The method according to claim 1 , wherein in the rating step uses a classification based on graph embedding.
5 . The method according to claim 4 , wherein the classification comprises: future knowledge graph forecasting, graph embedding and classification.
6 . The method according to claim 1 , wherein in the rating step, one classifier is used for each objective of the multiple different objectives.
7 . The method according to claim 1 , wherein in the considering step uses a multi-objective optimization.
8 . The method according to claim 7 , wherein the multi-objective optimization comprises predicting at least one future label of each decision.
9 . The method according to claim 8 , wherein the at least one future label of each decision is regarding a definable point in time or definable time steps.
10 . The method according to claim 7 , wherein the multi-objective optimization comprises the use of a Pareto-front selector, and wherein the decision is provided on the basis of a Pareto-front selection.
11 . The method according to claim 10 , wherein the Pareto-front selector incorporates a forecast uncertainty in an additional factor and trains its weights on previous cases from a database.
12 . The method according to claim 10 , wherein the Pareto-front selector provides the steps: Pareto-graph creation and decision-making among Pareto-options.
13 . The method according to claim 12 , wherein a future label forecasting is provided.
14 . The method according to claim 1 , wherein the decision is provided by an explanation module, and wherein the explanation module is adapted to interact with a user.
15 . The method according to claim 1 , wherein the temporal knowledge graph is updated as soon as new information is available.
16 . The method according to claim 1 , wherein the environment includes public domains, public safety, public services, governmental institutions, and/or health care.
17 . The method according to claim 16 , wherein the public domain environment comprises a control room or law enforcement and traffic monitoring and adjustment.
18 . The method according to claim 16 , wherein the public services environment comprises a job center for providing intelligent assignment and routing, and wherein governmental institutions environment comprises a smart city with a smart grid system.
19 . The method according to claim 16 , wherein the healthcare environment comprises drug development, or the biomedical field.
20 . A data processing system for decision-making regarding a decision in an environment, wherein decision-making is performed in view of multiple different objectives, the data processing system comprising:
collecting means for collecting information within the environment;
describing means for describing the information in at least one temporal knowledge graph (TKG);
forecasting means for forecasting a future development of one future state or more future states at a future time or more future points in time, under different decisions by the at least one temporal knowledge graph;
describing means for describing each resulting future state/decision combination by a corresponding temporal knowledge graph;
rating means for rating an adherence of each forecasted future state to each objective of the multiple different objectives;
considering means for considering a trade-off between the different objectives for decision-making in a time-aware manner; and
providing means for providing the decision.