The dynamic state context of the decision-making system is generated in the present moment.

Abstract

A dynamic state context generation system for a decision-making system includes a transmission interface, a computer-readable storage medium, and a processor electrically connected to the former two. The system loads subject characteristic data and extracts behavioral and event data falling within a specified range using a sliding time window. After temporal alignment, the data is mapped to corresponding fields in a multi-dimensional structure to form integrated data. This integrated data, along with the structure definition, is serialized and encapsulated in real-time to generate a state context snapshot. The state context snapshot is then converted into either a vector format or a structured data format, allowing it to be provided as input to either an artificial intelligence model or a rule engine for real-time decision-making. This achieves the technical benefits of improving the accuracy of real-time decision-making and cross-system reusability.

Claims

1. A dynamic state context generation system for a decision-making system, comprising: a transmission interface for receiving multiple streams of data continuously generated by an event stream, wherein, Each of the multiple streaming data entries includes a timestamp; a computer-readable storage medium is used to store multiple computer-executable instructions and a subject characteristic data; and a processor is used to be electrically connected to the transmission interface and the computer-readable storage medium, the processor executing the multiple computer-executable instructions to be configured to: load the subject characteristic data from the computer-readable storage medium, and set a sliding time window, the sliding time window covering a range before a current time point and within a predetermined time interval; extract from the multiple streaming data entries received by the transmission interface those whose timestamps fall within the sliding time window, as behavioral data and event data corresponding to the current time point; perform time-series alignment of the behavioral data and the event data according to the timestamps, and map the time-series aligned behavioral data and the event data, together with the loaded subject characteristic data, to a corresponding field in a multi-dimensional structure, to integrate them into integrated data corresponding to the current time point; The integrated data, along with one of the structural definitions of the multi-dimensional structure, is serialized to instantly encapsulate and generate a state context snapshot corresponding to the current point in time; the state context snapshot is converted into one of a vector format and a structured data format to generate a formatted state context snapshot; and, depending on the format of the formatted state context snapshot, the formatted state context snapshot is selected to be provided to one of an artificial intelligence model and a rule engine as input for an instant decision.

2. As in the dynamic state context generation system of the decision system of request item 1, when the processor integrates the integrated data, the processor is further configured to parse the validity of the behavioral data and the event data at the current time point based on the timestamp, and only associate the data that is determined to be valid with the field of the multi-dimensional structure.

3. A dynamic state context generation system for the decision system of Request 1, wherein the formatted state context snapshot is in the vector format, and when the processor converts the state context snapshot into the vector format, the processor is further configured to input the content of the state context snapshot into an embedding model to generate a vector representation corresponding to the state context snapshot and having a predetermined dimension, and to use the vector representation as the formatted state context snapshot.

4. A dynamic state context generation system for the decision system of Request 1, wherein the processor is further configured to: when the formatted state context snapshot is in the vector format, input the formatted state context snapshot into the artificial intelligence model to perform a decision inference to obtain a decision output result; and when the formatted state context snapshot is in the structured data format, input the formatted state context snapshot into the rule engine to perform a rule comparison to obtain a rule output result.

5. A dynamic state context generation system for the decision system of Request 1, wherein the processor is further configured to: when generating the state context snapshot, retain the integrated data corresponding to a previous point in time as a previous integrated data; when the event stream generates new streaming data, determine, based on the new streaming data, at least one field among the multiple fields of the multi-dimensional structure affected by the new streaming data, and perform an update only on the content of the at least one field in the previous integrated data, and remove the content of the multiple streaming data in the previous integrated data that falls outside the sliding time window, to generate the integrated data corresponding to a new current point in time; and serialize the integrated data corresponding to the new current point in time to encapsulate and generate the state context snapshot corresponding to the new current point in time.

Images