Learning management system and learning management method

• US 11,475,373 B2
• Filed: 10/30/2019
• Issued: 10/18/2022
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Ito, Kiyoto, Fuji, Taiki, Esaki, Kanako, Matsumoto, Kohsei
• Primary Examiner(s): Ahn, Sam K
• Assistant Examiner(s): Singh, Amneet
• Application Number:
  US16/668,892
• Publication Number:
  US 20200134509A1
• Time in Patent Office: 1,084 Days
• CPC Class Codes: G06F 17/15 Correlation function computation including computation of convolution operations arithmetic circuits for sum of products per se, e.g. multiply-accumulators G06F7/5443; digital filters, e.g. FIR, IIR, adaptive filters H03H17/00
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06Q 10/08 Logistics, e.g. warehousing, loading or distribution; Inventory or stock management

Abstract

A retail agent determines a parameter of an activity proposal model by using data stored in a past record database, and determines parameters of an activity determination model and an activity value evaluation model by further using base activity simulation data. Consequently, it is possible to appropriately determine parameters of a subsystem control method in a complex system which cannot be embodied as a simulator and shows a significant change with respect to past record data.

Claims

1. A learning management system which learns a behavior of a predetermined control target, the learning management system comprising:

a past record database which accumulates past record data of the predetermined control target; and

a server including a processor and a memory, wherein the processor is configured to execute programmed instructions that, when executed by the processor, cause the server to

store a first correlation model of a record learning portion of the server, said learning portion being configured to cause the processor to execute instructions comprising the first correlation model which causes the server output measure proposal data from data related to the behavior of the predetermined control target;

store a second correlation model of an activity learning portion of the server, said activity learning portion being configured to cause the processor to execute instructions comprising the second correlation model which causes the server to output measure data from simulation state data and the measure proposal data, the simulation state data simulating the behavior of the predetermined control target;

store a third correlation model of the activity learning portion of the server, said activity learning portion being configured to cause the processor to execute instructions comprising the third correlation model which outputs a measure evaluation value from the simulation state data, the measure proposal data, and the measure data;

determine a parameter of the first correlation model, using a first record learning portion of the record learning portion configured to calculate a correlation between the data related to the behavior of the predetermined control target, and the measure proposal data based on a predetermined evaluation value and the past record data stored in the past record database;

obtain the simulation state data, using the record learning portion and the activity learning portion of the server configured to simulate the behavior of the predetermined control target;

generate the measure proposal data by inputting the obtained simulation state data to the first correlation model, and, using the record learning portion and the activity learning portion of the server configured to generate the measure data by inputting the simulation state data obtained and the measure proposal data to the second correlation model;

determine parameters of the second correlation model and the third correlation model, using the activity learning portion of the server, based on the simulation state data obtained, the measure proposal data generated, the measure data, and a predetermined evaluation logic; and

determine the parameter of the first correlation model again, using a first record learning portion of the record learning portion configured to calculate the correlation between the data related to the behavior of the predetermined control target and the measure proposal data based on an evaluation value obtained by the third correlation model, and the past record data stored in the past record database.

2. The learning management system according to claim 1, further comprising a simulation activity database which associates and stores the simulation state data, and the measure proposal data and the measure data.

3. The learning management system according to claim 1, wherein the simulation state data is newly obtained based on an adjustment amount of the parameter of the first correlation model determined again.

4. A learning management method for learning a behavior of a predetermined control target, the method comprising:

accumulating past record data of the predetermined control target using a record database;

storing a first correlation model which outputs measure proposal data from data related to the behavior of the predetermined control target;

storing a second correlation model which outputs measure data from simulation state data and the measure proposal data, the simulation state data simulating the behavior of the predetermined control target;

storing a third correlation model which outputs a measure evaluation value from the simulation state data, the measure proposal data, and the measure data;

determining, in a first record learning step, a parameter of the first correlation model by calculating a correlation between the data related to the behavior of the predetermined control target, and the measure proposal data based on a predetermined evaluation value and the past record data stored in the record database;

obtaining, in a simulation state data obtaining step, the simulation state data which simulates the behavior of the predetermined control target;

generating, in a measure data generating step, the measure proposal data by inputting the simulation state data obtained in the simulation state data obtaining step to the first correlation model, and generating the measure data by inputting the simulation state data obtained in the simulation state data obtaining step and the measure proposal data to the second correlation model;

determining, in an activity learning step, parameters of the second correlation model and the third correlation model based on the simulation state data obtained in the simulation state data obtaining step, the measure proposal data generated in the measure data generating step, the measure data generated in the measure data generating step, and a predetermined evaluation logic; and

determining, in a second record learning step, the parameter of the first correlation model again by calculating the correlation between the data related to the behavior of the predetermined control target and the measure proposal data based on an evaluation value obtained by the third correlation model, and the past record data stored in the record database.