# US 12,602,638 B2 · Risk management system and risk management method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,602,638 B2
- Document ID: US12602638B2
- Status: Granted
- Kind: B2
- Application number: 18/341,035
- Filed: 2023-06-26
- Published: 2023-10-19
- Granted: 2026-04-14
- Earliest claimed: 2021-01-07
- Assignees: Mitani Sangyo Co., Ltd.; Yasushi Kiyoki
- Applicants: Mitani Sangyo Co., Ltd.; Yasushi Kiyoki
- Inventors: Yasushi Kiyoki; Yukio Chen; Koichi Mori; Toshiaki Fukabori; Tomokazu Shigeri; Yoshikane Oda
- Web page: https://patentgrep.com/patent/US12602638B2

## CPC

- G06Q 10/0635: Risk analysis of enterprise or organisation activities
- G06Q 10/04: Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202)
- G06Q 10/06: Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

A risk management system, which is a risk management system of upper layer system connected with a plurality of types of lower layer systems, includes an acquisition unit that acquires first sensing data acquired by a first lower layer system capable of acquiring first type data and second sensing data acquired by a second lower layer system capable of acquiring data of a different type from first sensing data; a normalization processing unit that generates integrated normalized data by performing arithmetic processing for generating first normalized data based on the first sensing data and generating second normalized data based on the second sensing data, and performing first arithmetic process for generating integrated normalized data by performing second arithmetic process for generating first normalized data and second normalized data as one piece of data; and a prediction processing unit that generates predicted risk data based on the integrated normalized data.

## Claims

### Claim 1

A risk management system for improving prediction accuracy of labor-related risk assessment through adaptive machine learning, the risk management system being an upper layer system connected to a plurality of lower layer systems, the risk management system comprising:
a first processor; and
a first memory device configured to store a program executed by the first processor to cause the first processor to:
acquire, from a first lower layer system comprising a first sensor type connected to a second processor and a second memory device, first sensing data comprising heat index measurements;
acquire, from a second lower layer system comprising a second sensor type, different from the first sensor type, and connected to a third processor and a third memory device, second sensing data comprising workload measurements;
train a machine learning model to predict risk data using historical first sensing data, historical second sensing data, and historical risk data serving as teacher data, the historical data being acquired in advance;
acquire user-specific information related to a human worker, the user-specific information including two or more of a length of service, environment information, age, and sex;
generate user-adapted first normalized data by:
determining a first normalized reference value based on the acquired user-specific information; and
applying a first normalization arithmetic formula that transforms the first sensing data relative to the first normalized reference value, wherein the transformation adjusts the heat index measurements in the acquired first sensing data to account for individual worker characteristics;
generate user-adapted second normalized data by:
determining a second normalized reference value based on the acquired user-specific information; and
applying a second normalization arithmetic formula that transforms the second sensing data relative to the second normalized reference value, wherein the transformation adjusts the workload measurements in the acquired second sensing data to account for the individual worker characteristics;
synthesize the user-adapted first normalized data and the user-adapted second normalized data into integrated normalized data through an arithmetic process that generates one coherent data for input into the machine learning model;
generate improved predicted risk data by inputting the integrated normalized data into the trained machine learning model, wherein the predicted risk data exhibits increased accuracy through the user-specific normalization that adapts the machine learning model inputs to individual worker characteristics; and
periodically retrain the trained machine learning model after an elapsed period of time using updated user-specific information, newly acquired first sensing data, newly acquired second sensing data, and actual risk outcomes from the elapsed period of time as new teacher data,
wherein the machine learning model generates the predicted risk data by assigning a weight value to at least one of the first normalized data, the second normalized data and the integrated normalized data based on a predetermined condition, and
wherein the predicted risk data is adjusted based on the weight value.

### Claim 2 (depends on 1)

The risk management system according to claim 1, wherein
the first normalization method is set based on first identification information associated with the first sensing data; and
the second normalization method is set based on second identification information associated with the second sensing data.

### Claim 3 (depends on 1)

The risk management system according to claim 1, wherein
the predetermined condition is set based on the user information.

### Claim 4 (depends on 1)

The risk management system according to claim 1, wherein
the program causes the first processor to:
transmit warning information to a user terminal in accordance with the predicted risk data.

### Claim 5 (depends on 1)

The risk management system according to claim 1, wherein the user-adapted first normalized data is generated such that:
in a range equal to or less than a first threshold value, warning level data and the normalized data are set to a minimum value indicating negligible risk,
in a range greater than the first threshold value and equal to or less than a second threshold value, the warning level data and the normalized data are set to increase by a first increment rate, respectively, and
in a range greater than the second threshold value and equal to or less than a third threshold value, the warning level data and the normalized data are set to increase at a second increment rate greater than the first increment rate, thereby providing nonlinear risk assessment that reflects increasing danger at higher heat indices.

### Claim 6 (depends on 1)

The risk management system according to claim 1, wherein the first processor further executes the program to generate warning information for notifying a risk for future work based on the improved predicted risk data, wherein the improved predicted risk data accounts for both environmental conditions including heat index factors and user-specific information to enable risk management according to each user.

### Claim 7

A risk management method for improving prediction accuracy of labor-related risk assessment through adaptive machine learning, the risk management method comprising:
acquiring, from a first lower layer system comprising a first sensor type, first sensing data comprising heat index measurements;
acquiring, from a second lower layer system comprising a second sensor type, different from the first sensor type, second sensing data comprising workload measurements;
training a machine learning model to predict risk data using historical first sensing data, historical second sensing data, and historical risk data serving as teacher data, the historical data being acquired in advance;
acquiring user-specific information related to a human worker, the user-specific information including two or more of a length of service, environment information, age, and sex;
generating user-adapted first normalized data by:
determining a first normalized reference value based on the acquired user-specific information; and
applying a first normalization arithmetic formula that transforms the first sensing data relative to the first normalized reference value, wherein the transformation adjusts the heat index measurements in the acquired first sensing data to account for individual worker characteristics;
generating user-adapted second normalized data by:
determining a second normalized reference value based on the acquired user-specific information; and
applying a second normalization arithmetic formula that transforms the second sensing data relative to the second normalized reference value, wherein the transformation adjusts the workload measurements in the acquired second sensing data to account for the individual worker characteristics;
synthesizing the user-adapted first normalized data and the user-adapted second normalized data into integrated normalized data through an arithmetic process that generates one coherent data for input into the machine learning model;
generating improved predicted risk data by inputting the integrated normalized data into the trained machine learning model, wherein the predicted risk data exhibits increased accuracy through the user-specific normalization that adapts the machine learning model inputs to individual worker characteristics; and
periodically retraining the trained machine learning model after an elapsed period of time using updated user-specific information, newly acquired first sensing data, newly acquired second sensing data, and actual risk outcomes from the elapsed period of time as new teacher data,
wherein the machine learning model generates the predicted risk data by assigning a weight value to at least one of the first normalized data, the second normalized data and the integrated normalized data based on a predetermined condition, and
wherein the predicted risk data is adjusted based on the weight value.

### Claim 8 (depends on 7)

The risk management method according to claim 7, wherein
the first normalization method is set based on first identification information associated with the first sensing data; and
the second normalization method is set based on second identification information associated with the second sensing data.

### Claim 9 (depends on 7)

The risk management method according to claim 7, wherein
the predetermined condition is set based on the user information.

### Claim 10 (depends on 7)

The risk management method according to claim 7 further comprising:
transmitting warning information to a user terminal in accordance with the predicted risk data.