# US 12,008,070 B2 · Method and apparatus for predicting time-series data

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

- Number: US 12,008,070 B2
- Document ID: US12008070B2
- Status: Granted
- Kind: B2
- Application number: 18/344,187
- Filed: 2023-06-29
- Published: 2023-10-26
- Granted: 2024-06-11
- Earliest claimed: 2020-01-17
- Assignees: Samsung Electronics Co., Ltd.
- Applicants: Samsung Electronics Co., Ltd.
- Inventors: Jooyoo Kim; Hyungseon Song; Kyungjae Kim; Chandra Ashok Maloo; Dongjun Shin; Pavan Kumar
- Web page: https://patentgrep.com/patent/US12008070B2

## CPC

- G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)}
- F24F 11/47: Responding to energy costs
- F24F 11/56: Remote control
- F24F 11/63: Electronic processing
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- H04L 67/125: involving control of end-device applications over a network
- F24F 2140/60: Energy consumption
- G06F 2218/12: Classification; Matching
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- F24F: AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING (removing dirt or fumes from areas where they are produced B08B15/00; vertical ducts for carrying away waste gases from buildings E04F17/02; tops for chimneys or ventilating shafts, terminals for flues F23L17/02)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

A time-series data prediction method is provided. The method includes storing a time-series model, inputting time-series data to the time-series model, and outputting a plurality of pieces of prediction data which is a preset number of pieces of prediction data or more based on the time-series model. The inputting of the time-series data to the time-series model includes substituting invalid data included in the time-series data with preset valid data, and inputting the time-series data including the valid data to the time-series model.

## Claims

### Claim 1

A time-series data prediction method comprising:
collecting, by a processor, a first time-series data including a power consumption data for a preset first time period of an air conditioner;
inputting, by the processor, the first time-series data including invalid data into a time-series model stored in a memory to obtain a first plurality of prediction data including a prediction data corresponding to the invalid data;
replacing, by the processor, the invalid data included in the first time-series data with preset valid data;
inputting, by the processor, the first time-series data including the valid data into the time-series model to obtain a second plurality of prediction data including a prediction data corresponding to the replaced valid data;
obtaining, by the processor, a first similarity between the first plurality of prediction data and a plurality of reference data and a second similarity between the plurality of second plurality of prediction data and the plurality of reference data;
collecting, by the processor, a second time-series data including a power consumption data for a second time period after the first time period of the air conditioner; and
determining, by the processor, whether to replace invalid data included in the second time-series data with a preset valid data by comparing the first similarity with the second similarity,
wherein the invalid data is null data corresponding to power consumption less than or equal to a preset reference value, and the valid data is data other than the null data.

### Claim 2 (depends on 1)

The time-series data prediction method of claim 1, wherein the determining of whether to replace invalid data included in the second time-series data with a preset valid data comprises:
replacing the invalid data included in the second time-series data with the preset valid data based on the second similarity greater than the first similarity.

### Claim 3 (depends on 2)

The time-series data prediction method of claim 2, further comprises:
inputting, by the processor, the second time-series data including the replaced valid data into the time-series model.

### Claim 4 (depends on 3)

The time-series data prediction method of claim 3, further comprises:
obtaining, by the processor, a third plurality of prediction data including a prediction data corresponding to the replaced valid data of the second time-series data as outputs of the time-series model; and
restoring, by the processor, the prediction data corresponding to the replaced valid data among the third plurality of prediction data into the invalid data.

### Claim 5 (depends on 4)

The time-series data prediction method of claim 4, wherein the restoring of the prediction data corresponding to the replaced valid data among the third plurality of prediction data into the invalid data comprises:
restoring the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data based on a period of the invalid data included in the second time-series data.

### Claim 6 (depends on 4)

The time-series data prediction method of claim 4, wherein the restoring of the prediction data corresponding to the replaced valid data among the third plurality of prediction data into the invalid data comprises:
restoring the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data through a probability distribution based on a number of pieces of the invalid data included in the second time-series data.

### Claim 7 (depends on 1)

The time-series data prediction method of claim 1, wherein the time-series model includes at least one of Prophet, Autoregressive Integrated Moving Average (ARIMA), Recurrent Neural Networks (RNN), Long Short-Term Memory models (LSTM), or Sequence-to-Sequence (Seq2Seq).

### Claim 8

An apparatus comprising:
a memory storing a time-series model; and
a processor configured to:
collect a first time-series data including a power consumption data for a preset first time period of an air conditioner,
input the first time-series data including invalid data into a time-series model to obtain a first plurality of prediction data including a prediction data corresponding to the invalid data,
replace the invalid data included in the first time-series data with preset valid data,
input the first time-series data including the valid data into the time-series model to obtain a second plurality of prediction data including a prediction data corresponding to the replaced valid data,
obtain a first similarity between the first plurality of prediction data and a plurality of reference data and a second similarity between the plurality of second plurality of prediction data and the plurality of reference data,
collect a second time-series data including a power consumption data for a second time period after the first time period of the air conditioner, and
determine whether to replace invalid data included in the second time-series data with a preset valid data by comparing the first similarity with the second similarity,
wherein the invalid data is null data corresponding to power consumption less than or equal to a preset reference value, and the valid data is data other than the null data.

### Claim 9 (depends on 8)

The apparatus of claim 8, wherein the processor is further configured to:
replace the invalid data included in the second time-series data with the preset valid data based on the second similarity greater than the first similarity.

### Claim 10 (depends on 9)

The apparatus of claim 9, wherein the processor is further configured to:
input the second time-series data including the replaced valid data into the time-series model.

### Claim 11 (depends on 10)

The apparatus of claim 10, wherein the processor is further configured to:
obtain a third plurality of prediction data including a prediction data corresponding to the replaced valid data of the second time-series data as outputs of the time-series model, and
restore the prediction data corresponding to the replaced valid data among the third plurality of prediction data into the invalid data.

### Claim 12 (depends on 11)

The apparatus of claim 11, wherein the processor is further configured to:
restore the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data based on a period of the invalid data included in the second time-series data.

### Claim 13 (depends on 11)

The apparatus of claim 11, wherein the processor is further configured to:
restore the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data through a probability distribution based on a number of pieces of the invalid data included in the second time-series data.

### Claim 14

A system comprising:
an air conditioner; and
a server configured to communicate with the air conditioner,
wherein the server is configured to:
collect a first time-series data including a power consumption data for a preset first time period of the air conditioner,
input the first time-series data including invalid data into a time-series model to obtain a first plurality of prediction data including a prediction data corresponding to the invalid data,
replace the invalid data included in the first time-series data with preset valid data,
input the first time-series data including the valid data into the time-series model to obtain a second plurality of prediction data including a prediction data corresponding to the replaced valid data,
obtain a first similarity between the first plurality of prediction data and a plurality of reference data and a second similarity between the plurality of second plurality of prediction data and the plurality of reference data,
collect a second time-series data including a power consumption data for a second time period after the first time period of the air conditioner, and
determine whether to replace invalid data included in the second time-series data with a preset valid data by comparing the first similarity with the second similarity,
wherein the invalid data is null data corresponding to power consumption less than or equal to a preset reference value, and the valid data is data other than the null data.

### Claim 15 (depends on 14)

The system of claim 14, wherein the server is further configured to:
replace the invalid data included in the second time-series data with the preset valid data based on the second similarity greater than the first similarity.

### Claim 16 (depends on 15)

The system of claim 15, wherein the server is further configured to:
input the second time-series data including the replaced valid data into the time-series model.

### Claim 17 (depends on 16)

The system of claim 16, wherein the server is further configured to:
obtain a third plurality of prediction data including a prediction data corresponding to the replaced valid data of the second time-series data as outputs of the time-series model,
restore the prediction data corresponding to the replaced valid data among the third plurality of prediction data into the invalid data, and
transmit the third plurality of prediction data including the restored invalid data to the air conditioner.

### Claim 18 (depends on 17)

The system of claim 17, wherein the server is further configured to:
restore the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data based on a period of the invalid data included in the second time-series data.

### Claim 19 (depends on 17)

The system of claim 17, wherein the server is further configured to:
restore the prediction data being a part of the third plurality of prediction data output from the time-series model to the invalid data through a probability distribution based on a number of pieces of the invalid data included in the second time-series data.