# US 2026/0094054 A1 · Automated machine learning based workflow for timeseries forecasting

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

- Number: US 2026/0094054 A1
- Document ID: US20260094054A1
- Status: Published application
- Kind: A1
- Application number: 18/901,722
- Filed: 2024-09-30
- Published: 2026-04-02
- Earliest claimed: 2024-09-30
- Assignees: Fujitsu Limited; Archimedes Controls Corp.
- Applicants: Fujitsu Limited
- Inventors: Michael McTHROW; Kanji UCHINO; Wenli YU; Liangcai TAN
- Web page: https://patentgrep.com/patent/US20260094054A1

## CPC

- G05B 23/0243: {model based detection method, e.g. first-principles knowledge model}
- G06N 20/00: Machine learning
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 5/04: Inference or reasoning models
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

In an embodiment, workflow for timeseries forecasting may be performed based on automated machine learning. Sensor data for measurement parameter is received from plurality of sensors installed in built environment and the received sensor data is stored in table of relational database. Cut-off record associated with previous training checkpoint is determined of the forecasting model for the measurement parameter. Records including new records are determined for which respective timestamps occur after the measurement timestamp of cut-off record. Size of the determined records are compared with threshold size and training dataset is prepared. The forecasting model is trained on the training dataset based on the comparison.

## Claims

### Claim 1

A method, executed by at least one processor, comprising:
receiving sensor data for a measurement parameter from a sensor of a plurality of sensors installed in a built environment;
storing the received sensor data as new records in a table of a relational database;
determining a cut-off record associated with a previous training checkpoint of a forecasting model for the measurement parameter;
determining, from the table, records including the new records for which respective measurement timestamps, as specified in the table, occur after a measurement timestamp of the cut-off record;
comparing a size of the determined records with a threshold size;
preparing a training dataset based on the determined records; and
training the forecasting model on the training dataset based on the comparison.

### Claim 2 (depends on 1)

The method according to claim 1, wherein the built environment includes one of a data center and an industrial facility.

### Claim 3 (depends on 1)

The method according to claim 1, wherein the forecasting model is trained on the training dataset based on a determination that the size of the records is above the threshold size.

### Claim 4 (depends on 1)

The method according to claim 1, further comprising:
preparing an input for the forecasting model based on the received sensor data and a determination that the size of the records is below the threshold size;
applying the forecasting model on the prepared input to generate a forecast value of the measurement parameter for a future timestamp; and
querying a knowledge database based on the forecast value to determine a suggestion for equipment installed in the built environment.

### Claim 5 (depends on 4)

The method according to claim 4, wherein the suggestion includes an action to perform a repair or maintenance of the equipment, a servicing of the equipment, or a replacement of the equipment.

### Claim 6 (depends on 4)

The method according to claim 4, wherein the suggestion indicates whether the forecast value corresponds to a faulty state of the equipment.

### Claim 7 (depends on 4)

The method according to claim 4, further comprising controlling a display device associated with an administrator of the built environment to display the forecast value along with the suggestion.

### Claim 8 (depends on 4)

The method according to claim 4, wherein the knowledge database comprises:
conditions that include a plurality of historical values of the measurement parameter and a respective plurality of timestamps associated with the plurality of historical values,
decision information that include a plurality of suggestions corresponding to the plurality of historical values, and
a source type associated with the conditions and the decision information.

### Claim 9 (depends on 8)

The method according to claim 8, wherein the querying of the knowledge database comprises:
determining, from the plurality of historical values, a historical value that matches the forecast value; and
retrieving the suggestion that corresponds to the historical value from the plurality of suggestions.

### Claim 10 (depends on 1)

The method according to claim 1, wherein the preparing the training dataset comprises:
extracting, from the records of the table, a feature column that stores values of the measurement parameter and the respective measurement timestamps;
sorting the values in the feature column based on the respective measurement timestamps;
computing a median interval between the respective measurement timestamps;
executing, after the sorting, an aggregation query on the feature column to generate an aggregated feature column; and
determining missing measurement timestamps in the aggregated feature column;
filling missing values corresponding to the missing measurement timestamps in the aggregated feature column;
determining a sliding window size for the aggregated feature column; and
obtaining the training dataset from the aggregate feature column based on the sliding window size.

### Claim 11 (depends on 10)

The method according to claim 10, wherein the execution of the aggregation query comprises:
dividing the respective measurement timestamps by the computed median interval to determine a plurality of aggregate intervals;
selecting, from the feature column, a set of values corresponding to each unique aggregate interval of the plurality of aggregate intervals;
computing an average value of the set of values; and
grouping the average value based on the plurality of aggregate intervals, wherein the aggregate feature column includes the average value corresponding to each unique aggregate interval of the plurality of aggregate intervals.

### Claim 12 (depends on 1)

The method according to claim 1, wherein the forecasting model is trained using an automated machine learning operation.

### Claim 13

One or more non-transitory computer-readable storage media configured to store instructions that, in response to being executed, cause a system to perform operations, the operations comprising:
receiving sensor data for a measurement parameter from a sensor of a plurality of sensors installed in a built environment;
storing the received sensor data as new records in a table of a relational database;
determining a cut-off record associated with a previous training checkpoint of a forecasting model for the measurement parameter;
determining, from the table, records including the new records for which respective measurement timestamps, as specified in the table, occur after a measurement timestamp of the cut-off record;
comparing a size of the determined records with a threshold size;
preparing a training dataset based on the determined records; and
training the forecasting model on the training dataset based on the comparison.

### Claim 14 (depends on 13)

The one or more non-transitory computer-readable storage media according to claim 13, wherein the forecasting model is trained on the training dataset based on a determination that the size of the records is above the threshold size.

### Claim 15 (depends on 13)

The one or more non-transitory computer-readable storage media according to claim 13, wherein the operations further comprise:
preparing an input for the forecasting model based on the received sensor data and a determination that the size of the records is below the threshold size;
applying the forecasting model on the prepared input to generate a forecast value of the measurement parameter for a future timestamp; and
querying a knowledge database based on the forecast value to determine a suggestion for equipment installed in the built environment.

### Claim 16 (depends on 15)

The one or more non-transitory computer-readable storage media according to claim 15, wherein the operations further comprise controlling a display device associated with an administrator of the built environment to display the forecast value along with the suggestion.

### Claim 17 (depends on 15)

The one or more non-transitory computer-readable storage media according to claim 15, wherein the knowledge database comprises:
conditions that include a plurality of historical values of the measurement parameter and a respective plurality of timestamps associated with the plurality of historical values,
decision information that include a plurality of suggestions corresponding to the plurality of historical values, and
a source type associated with the conditions and the decision information.

### Claim 18 (depends on 13)

The one or more non-transitory computer-readable storage media according to claim 13, wherein the preparing of the training dataset comprises:
extracting, from the records of the table. a feature column that stores values of the measurement parameter and the respective measurement timestamps;
sorting the values in the feature column based on the respective measurement timestamps;
computing a median interval between the respective measurement timestamps;
executing, after the sorting, an aggregation query on the feature column to generate an aggregated feature column; and
determining missing measurement timestamps in the aggregated feature column;
filling missing values corresponding to the missing measurement timestamps in the aggregated feature column;
determining a sliding window size for the aggregated feature column; and
obtaining the training dataset from the aggregate feature column based on the sliding window size.

### Claim 19 (depends on 18)

The one or more non-transitory computer-readable storage media according to claim 18, wherein the execution of the aggregation query comprises:
dividing the respective measurement timestamps by the computed median interval to determine a plurality of aggregate intervals;
selecting, from the feature column, a set of values corresponding to each unique aggregate interval of the plurality of aggregate intervals;
computing an average value of the set of values; and
grouping the average value based on the plurality of aggregate intervals, wherein the aggregate feature column includes the average value corresponding to each unique aggregate interval of the plurality of aggregate intervals.

### Claim 20

A system, comprising:
a memory storing instructions; and
a processor, coupled to the memory, which executes the instructions to perform a process comprising:
receiving sensor data for a measurement parameter from a sensor of a plurality of sensors installed in a built environment;
storing the received sensor data as new records in a table of a relational database;
determining a cut-off record associated with a previous training checkpoint of a forecasting model for the measurement parameter;
determining, from the table, records including the new records for which respective measurement timestamps, as specified in the table, occur after a measurement timestamp of the cut-off record;
comparing a size of the determined records with a threshold size;
preparing a training dataset based on the determined records; and
training the forecasting model on the training dataset based on the comparison.
