# US 2021/0172918 A1 · SENSOR TUNING - SENSOR SPECIFIC SELECTION FOR IoT-ELECTRONIC NOSE APPLICATION USING GRADIENT BOOSTING DECISION TREES

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

- Number: US 2021/0172918 A1
- Document ID: US20210172918A1
- Status: Published application
- Kind: A1
- Application number: 16/707,679
- Filed: 2019-12-09
- Published: 2021-06-10
- Earliest claimed: 2019-12-09
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Mohammed Abdi; Aminat Adebiyi; Alberto Mannari; Andrea Fasoli; Ronald Robert Labby; Luisa Bozano; Pawan Chowdhary; Abubeker Abdullahi
- Web page: https://patentgrep.com/patent/US20210172918A1

## CPC

- G01N 33/0031: {comprising two or more sensors, e.g. a sensor array}
- G06F 18/2113: by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
- G06N 20/20: Ensemble learning
- G16Y 20/10: relating to the environment, e.g. temperature; relating to location
- G16Y 40/10: Detection; Monitoring
- G16Y 40/35: Management of things, i.e. controlling in accordance with a policy or in order to achieve specified objectives
- G06F 2218/08: Feature extraction
- G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16Y: INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]

## Abstract

Provided is a system and method for tuning an array of sensors to enable selection of the most suitable sensors for a target application. After extracting features from sensor raw data, the extracted features are ranked with gradient boosting decision trees to assign an importance value to each extracted feature. A threshold value for the entire set of extracted features is calculated and an importance score is calculated for the individual sensors of the array. Individual sensors with an importance score on or above the threshold value are selected for the target application.

## Claims

### Claim 1

A system comprising:
a sensor array; and
a processor with at least one algorithm stored therein,
wherein,
raw data obtained from the sensor array are input into the processor to form an input set of extracted features,
the at least one algorithm ranks the extracted features from the input set, wherein each extracted feature is given an importance value, and
an importance score for each sensor of the sensor array is determined based upon the ranking of the extracted features for each sensor.

### Claim 2 (depends on 1)

The system of claim 1, wherein a threshold value refines the input set, and individual sensors having an importance score on or above the threshold value are selected for a target application.

### Claim 3 (depends on 2)

The system of claim 2, wherein the threshold value Thres is calculated according to Formula 1:
Thres
=
∑
fi
♯


feat
Formula


1
wherein,
fi is the importance value for each extracted feature from the input set, and
#feat is the total number of extracted features from the input set.

### Claim 4 (depends on 1)

The system of claim 1, wherein the importance score ImpSn of a single sensor in the sensor array is calculated according to Formula 2:
ImpSn
=
∑
fis
♯


feats
Formula


2
wherein,
fis is the importance value for each extracted feature for a single sensor, and
#feats is the total number of extracted features for a single sensor.

### Claim 5 (depends on 1)

The system of claim 1, wherein the at least one algorithm comprises gradient boosting decision trees.

### Claim 6 (depends on 1)

The system of claim 1, wherein the sensor array is comprised of a plurality of gas sensors and the gas sensors measure volatile organic compounds (VOCs).

### Claim 7 (depends on 2)

The system of claim 2, wherein the target application is selected from the group consisting of air quality analyses, air pollution analyses, water pollution analyses, soil analyses, lab test, medical tests, tests for allergens, breathalyzer tests, food and beverage expiration date analyses, food and beverage spoilage analyses, alcohol content analyses, product authenticity, and combinations thereof.

### Claim 8 (depends on 7)

The system of claim 7, wherein the product authenticity comprises identification of a crypto-currency digital footprint embedded in the product.

### Claim 9

A system comprising:
a sensor array, and
a processor with at least one algorithm stored therein,
wherein,
raw data obtained from the sensor array are input into the processor to form an input set of extracted features, wherein each extracted feature is given an importance value,
the at least one algorithm ranks the extracted features from the input set, and
a threshold value comprising an aggregate value of the extracted features is established, wherein individual sensors of the sensor array having extracted features ranked on or above the threshold value are suitable for a target application.

### Claim 10 (depends on 9)

The system of claim 9, wherein the threshold value Thres is calculated according to Formula 1:
Thres
=
∑
fi
♯


feat
Formula


1
wherein,
fi is the importance value for each extracted feature from the input set, and
#feat is the total number of extracted features from the input set.

### Claim 11 (depends on 9)

The system of claim 9, wherein the extracted features of the individual sensors are ranked in relation to the threshold value with an importance score ImpSn that is calculated according to Formula 2:
ImpSn
=
∑
fis
♯


feats
Formula


2
wherein,
fis is the importance value for each extracted feature for a single sensor,
#feats is the total number of extracted features in for a single sensor.

### Claim 12 (depends on 11)

The system of claim 11, wherein an individual sensor with an importance score on or above the threshold level is suitable for the target application.

### Claim 13

A system comprising:
a sensor array; and
a processor with at least one algorithm stored therein,
wherein,
raw data obtained from the sensor array are input into the processor to form an input set of extracted features,
the at least one algorithm ranks extracted features from the input set, wherein each extracted feature is given an importance value,
a threshold value comprising an aggregate value of the extracted features is established, and
an importance score for each sensor of the sensor array is determined based upon the ranking of the extracted features for each sensor.

### Claim 14 (depends on 13)

The system of claim 13, wherein individual sensors of the sensor array having an importance score on or above the threshold value are suitable for a target application.

### Claim 15 (depends on 13)

The system of claim 13, wherein the threshold value Thres is calculated according to Formula 1:
Thres
=
∑
fi
♯


feat
Formula


1
wherein,
fi is the importance value for each extracted feature from the input set, and
#feat is the total number of extracted features in the input set.

### Claim 16 (depends on 13)

The system of claim 13, wherein the individual sensors having extracted features ranked on or above the threshold value are identified with an importance score that is calculated according to Formula 2:
ImpSn
=
∑
fis
♯


feats
Formula


2
wherein,
fis is the importance value for each extracted feature for a single sensor, and
#feats is the total number of features in a single sensor.

### Claim 17

A method comprising the steps of:
extracting features from an input set obtained from a sensor array;
ranking the extracted features with at least one algorithm, wherein each extracted feature is given an importance value;
calculating a threshold value for the ranked extracted features;
calculating an importance score for individual sensors of the sensor array based upon the ranking of the extracted features; and
selecting the individual sensors of the sensor array having an importance score on or above the threshold value for a target application.

### Claim 18 (depends on 17)

The method of claim 17, wherein the threshold value Thres is calculated according to Formula 1:
Thres
=
∑
fi
♯


feat
Formula


1
wherein,
fi is the importance value for each extracted feature from the input set, and
#feat is the total number of extracted features from the input set.

### Claim 19 (depends on 17)

The method of claim 17, wherein the importance score is calculated according to Formula 2:
ImpSn
=
∑
fis
♯


feats
Formula


2
wherein,
fis is the importance value for each extracted feature for a single sensor, and
#feats is the total number of extracted features extracted for a single sensor.

### Claim 20 (depends on 17)

The method of claim 17, wherein the feature values are ranked with an algorithm comprising gradient boosting decision trees.

### Claim 21 (depends on 17)

The method of claim 17, wherein the sensor array is comprised of a plurality of gas sensors and the gas sensors measure volatile organic compounds (VOCs).

### Claim 22 (depends on 17)

The method of claim 17, wherein the target application is selected from the group consisting of air quality analyses, air pollution analyses, water pollution analyses, soil analyses, lab test, medical tests, tests for allergens, breathalyzer tests, food and beverage expiration date analyses, food and beverage spoilage analyses, alcohol content analyses, product authenticity, and combinations thereof.

### Claim 23 (depends on 22)

The method of claim 22, wherein the product authenticity comprises identification of a crypto-currency digital footprint embedded in the product.

### Claim 24

A method comprising the steps of:
preparing a sensor array for a target application and applying information to the sensor array to form an input set;
extracting features from the input set and ranking the extracted features, wherein each extracted feature is given an importance value;
calculating a threshold value for the extracted features;
calculating an importance score for individual sensors of the sensor array based upon the ranking of the extracted features; and
selecting the individual sensors for the target application that have an importance score above the threshold value.

### Claim 25 (depends on 24)

The method of claim 24, wherein the threshold value Thres is calculated according to Formula 1:
Thres
=
∑
fi
♯


feat
Formula


1
wherein,
fi is the importance value for each extracted feature from the input set, and
#feat is the total number of extracted features from the input set.

### Claim 26 (depends on 24)

The method of claim 24, wherein the importance score ImpSn is calculated according to Formula 2:
ImpSn
=
∑
fis
♯


feats
Formula


2
wherein,
fis is the importance value for each extracted feature for a single sensor, and
#feats is the total number of extracted features in a single sensor.

### Claim 27 (depends on 24)

The method of claim 24, wherein the extracted features are ranked with an algorithm comprising gradient boosting decision trees.

### Claim 28 (depends on 24)

The method of claim 24, wherein the sensor array is comprised of a plurality of gas sensors and the gas sensors measure volatile organic compounds (VOCs).

### Claim 29 (depends on 24)

The method of claim 24, wherein the target application is selected from the group consisting of air quality analyses, air pollution analyses, water pollution analyses, soil analyses, lab tests, medical tests, tests for allergens, breathalyzer tests, food and beverage expiration date analyses, food and beverage spoilage analyses, alcohol content analyses, product authenticity, and combinations thereof.

### Claim 30 (depends on 29)

The method of claim 29, wherein the product authenticity comprises identification of a crypto-currency digital footprint embedded in the product.